home.social

Search

692 results for “ismina”

  1. bs2 @[email protected] ·

    > この辺りは尾鈴酸性岩からなり、地質上は尾鈴瀑布群の一部とも言えまが、「宮崎の自然と未来を守る会」さんにより、「石並川清布群」とされています。現状私が行った場所のみ掲載していますので、実際の数とは異なります。石並川水系の滝は、ほぼ「#宮崎の自然と未来を守る会 」の和田聖志さんが付けられた名前で載せていますが、地元では別の名で呼ばれているものもあるようです。

    miyahima.web.fc2.com/taki/miya

    #またたきの滝 #宮崎の滝 #日向の滝 #美しい滝
    #石並川水系 #石並川 #宮崎県日向市 #宮崎県日向市東郷
    #宮崎県日向市東郷町石並川

    #宮崎県日向市東郷町石並川 #宮崎県日向市東郷 #宮崎県日向市 #石並川 #石並川水系 #美しい滝
  2. 800mi @[email protected] ·

    „Das ist #SepidehGholian, seit Nov. 2018 immer wieder in der Haft der islamischen Republik. Sie wurde das 1. Mal festgenommen als sie die Arbeiter*innen Proteste unterstützt hat. Momentan in Evin, aber sie wurde in unterschiedlichen Gefängnissen festgehalten.
    Jina #MahsaAmini‌“

    nitter.sethforprivacy.com/khan
    #IranRevolution

    #iranrevolution #mahsaamini #sepidehgholian
  3. Jean Luc am Grimmsten @[email protected] ·

    Artikel-Eröffnung:

    "Herr Reker, die Ereignisse von Solingen, Magdeburg oder Aschaffenburg haben das Sicherheitsgefühl in der Bevölkerung verändert."

    Gleich mit einer herbeigeschriebenen Schimäre anfangen. Ihr Journalisten seid es doch, die das Thema mit hochgejazzt haben. Wir haben im Moment eine immer sicherere Gesellschaft (siehe zdf.de/nachrichten/panorama/kr) und ihr gebt den Schreihälsen die Bühne, nicht den Einordnern.

    *kotzen*

    nw.de/nachrichten/zwischen_wes €

    #kwalitätsdschornalismus #rechtsruck

    #kwalitatsdschornalismus #rechtsruck
  4. EshaHaber @[email protected] ·

    ABD basını bombayı patlattı! Trump, en yakın ismin üzerini çizdi! Görevden ayrılıyor: ABD medyasına yansıyan ve adı açıklanmayan Beyaz Saray kaynaklarına dayandırılan haberlere göre 15 Mart'ta ABD ordusunun Yemen'e düzenlediği saldırının planlarının sızdırılmasının başaktörlerinden ABD Ulusal Güvenlik Danışmanı Mike Waltz görevinden ayrılıyor.

    WALTZ GÖREVDEN AYRILIYOR

    İlk olarak CBS News kanalının… eshahaber.com.tr/haber/abd-bas EshaHaber.com.tr #ABD #Trump #Yemen #MikeWaltz #UlusalGüvenlik

    #abd #trump #yemen #mikewaltz #ulusalguvenlik
  5. People @[email protected] ·

    europesays.com/people/65034/ Putin Says Russia Ready For Talks After Long-Term Peace Deal With Ukraine #AccordingToReportsPublishedOnSunday #InternationalCommunity #PakitanAndIran #refugees #RussianPresidentVladimirPutinSaidHeWouldOnlyMeetUkrainianPresidentVolodymyrZelenskyyAfterA“sustainablePeaceAgreement”IsFinalizedBetweenTheTwoCountries #SECURITY #TalibanRegime #VolodymyrZelenskyy

    #volodymyrzelenskyy #talibanregime #security #russianpresidentvladimirputinsaidhewouldonlymeetukrainianpresidentvolodymyrzelenskyyaftera #refugees #pakitanandiran
  6. IT News @[email protected] ·

    Bug Eliminator Zaps with a Laser - Mosquitoes tend to be seen as an almost universal negative, at least in the lives ... - hackaday.com/2022/06/06/bug-el #machinelearning #neuralnetwork #laserhacks #neutralize #deterrent #mosquito #tracking #laser

    #laser #tracking #mosquito #deterrent #neutralize #laserhacks
  7. EshaHaber @[email protected] ·

    Picasso'nun bugüne kadar görülmemiş tablosu satıldı: Dünyaca ünlü İspanyol ressam Pablo Picasso'nun 80 yıl önce çizdiği tablo, ilk kez kamuoyunun önüne çıktı ve rekor fiyata satıldı. "Çiçek Şapkalı Kadın Büstü" ismini taşıyan tablo Paris'teki açık artırmada 27 milyon euro'ya alıcı buldu.

    Picasso'nun sevgilisi ve ilham perisi Dora Maar'ı tasvir ettiği tablonun dikkat çeken bir hikayesi… eshahaber.com.tr/haber/picasso EshaHaber.com.tr #Picasso #Sanat #Resim #AçıkArtırma #ÇiçekŞapkalıKadınBüstü

    #picasso #sanat #resim #acikartirma #ciceksapkalikadinbustu
  8. Maria Ismanah @[email protected] ·

    Our neighbor Aramis visiting - he is just gorgeous #Aramis #caturday #cat #feline

    #feline #cat #caturday #aramis
  9. Variety @[email protected] ·

    Christina Hendricks Teases ‘Good American Family,’ Based on ‘Mind-Blowing’ True Story, Admits ‘Every Actor in the World’ Wants to Be In ‘White Lotus’
    #Variety #Global #News #ChristinaHendricks #SeriesMania

    variety.com/2025/tv/global/chr

    #variety #global #news #christinahendricks #seriesmania
  10. Variety @[email protected] ·

    Christina Hendricks Teases ‘Good American Family,’ Based on ‘Mind-Blowing’ True Story, Admits ‘Every Actor in the World’ Wants to Be In ‘White Lotus’
    #Variety #Global #News #ChristinaHendricks #SeriesMania

    variety.com/2025/tv/global/chr

    #variety #global #news #christinahendricks #seriesmania
  11. Variety @[email protected] ·

    Christina Hendricks Teases ‘Good American Family,’ Based on ‘Mind-Blowing’ True Story, Admits ‘Every Actor in the World’ Wants to Be In ‘White Lotus’
    #Variety #Global #News #ChristinaHendricks #SeriesMania

    variety.com/2025/tv/global/chr

    #seriesmania #christinahendricks #news #global #variety
  12. Variety @[email protected] ·

    Christina Hendricks Teases ‘Good American Family,’ Based on ‘Mind-Blowing’ True Story, Admits ‘Every Actor in the World’ Wants to Be In ‘White Lotus’
    #Variety #Global #News #ChristinaHendricks #SeriesMania

    variety.com/2025/tv/global/chr

    #variety #global #news #christinahendricks #seriesmania
  13. EshaHaber @[email protected] ·

    Serdal Adalı, 3 kulübün ismini vererek MHK'yi topa tuttu: Sinir uçlarımızla oynuyor: Beşiktaş başkanı Serdal Adalı, açıklamalarda bulundu.

    Son maçlarda yaşanan hakem hatalarına vurgu yapan Adalı, Merkez Hakem Kurulu'na yönelik (MHK) sert ifadeler kullandı.

    "BEŞİKTAŞ, TRABZONSPOR VE RİZESPOR ÖZELİNDE DÖNEN BİR OYUN VAR"

    "Sorunun temel noktası TFF yönetim kurulu değil. Sorun Merkez Hakem Kurulu'nun… eshahaber.com.tr/haber/serdal- EshaHaber.com.tr #Beşiktaş #SerdalAdalı #MHK #Hakem #TFF

    #besiktas #serdaladali #mhk #hakem #tff
  14. EshaHaber @[email protected] ·

    Serdal Adalı, 3 kulübün ismini vererek MHK'yi topa tuttu: Sinir uçlarımızla oynuyor: Beşiktaş başkanı Serdal Adalı, açıklamalarda bulundu.

    Son maçlarda yaşanan hakem hatalarına vurgu yapan Adalı, Merkez Hakem Kurulu'na yönelik (MHK) sert ifadeler kullandı.

    "BEŞİKTAŞ, TRABZONSPOR VE RİZESPOR ÖZELİNDE DÖNEN BİR OYUN VAR"

    "Sorunun temel noktası TFF yönetim kurulu değil. Sorun Merkez Hakem Kurulu'nun… eshahaber.com.tr/haber/serdal- EshaHaber.com.tr #Beşiktaş #SerdalAdalı #MHK #Hakem #TFF

    #tff #hakem #mhk #serdaladali #besiktas
  15. EshaHaber @[email protected] ·

    Serdal Adalı, 3 kulübün ismini vererek MHK'yi topa tuttu: Sinir uçlarımızla oynuyor: Beşiktaş başkanı Serdal Adalı, açıklamalarda bulundu.

    Son maçlarda yaşanan hakem hatalarına vurgu yapan Adalı, Merkez Hakem Kurulu'na yönelik (MHK) sert ifadeler kullandı.

    "BEŞİKTAŞ, TRABZONSPOR VE RİZESPOR ÖZELİNDE DÖNEN BİR OYUN VAR"

    "Sorunun temel noktası TFF yönetim kurulu değil. Sorun Merkez Hakem Kurulu'nun… eshahaber.com.tr/haber/serdal- EshaHaber.com.tr #Beşiktaş #SerdalAdalı #MHK #Hakem #TFF

    #besiktas #serdaladali #mhk #hakem #tff
  16. EshaHaber @[email protected] ·

    Cumhurbaşkanı Erdoğan, Umman Sultanı ile görüştü: İsrail'in İran'a yönelik 13 Haziran'da başlattığı "Yükselen Aslan" saldırısı ve İran'ın misillemesi üçüncü gününde sürüyor. İsrail'e karşı İran da "Gerçek Vaat 3" ismini verdiği saldırıyı başlattı. Drone'ların yanı sıra yüzlerce balistik füzenin fırlatıldığı saldırıda Tel Aviv'de ve bazı İsrail kentlerinde patlamalar yaşandı.

    UMMAN SULTANI İLE… eshahaber.com.tr/haber/cumhurb EshaHaber.com.tr #Cumhurbaşkanı #Erdoğan #Umman #İran #İsrail

    #cumhurbaskani #erdogan #umman #iran #israil
  17. EshaHaber @[email protected] ·

    Cumhurbaşkanı Erdoğan, Umman Sultanı ile görüştü: İsrail'in İran'a yönelik 13 Haziran'da başlattığı "Yükselen Aslan" saldırısı ve İran'ın misillemesi üçüncü gününde sürüyor. İsrail'e karşı İran da "Gerçek Vaat 3" ismini verdiği saldırıyı başlattı. Drone'ların yanı sıra yüzlerce balistik füzenin fırlatıldığı saldırıda Tel Aviv'de ve bazı İsrail kentlerinde patlamalar yaşandı.

    UMMAN SULTANI İLE… eshahaber.com.tr/haber/cumhurb EshaHaber.com.tr #Cumhurbaşkanı #Erdoğan #Umman #İran #İsrail

    #israil #iran #umman #erdogan #cumhurbaskani
  18. EshaHaber @[email protected] ·

    Cumhurbaşkanı Erdoğan, Umman Sultanı ile görüştü: İsrail'in İran'a yönelik 13 Haziran'da başlattığı "Yükselen Aslan" saldırısı ve İran'ın misillemesi üçüncü gününde sürüyor. İsrail'e karşı İran da "Gerçek Vaat 3" ismini verdiği saldırıyı başlattı. Drone'ların yanı sıra yüzlerce balistik füzenin fırlatıldığı saldırıda Tel Aviv'de ve bazı İsrail kentlerinde patlamalar yaşandı.

    UMMAN SULTANI İLE… eshahaber.com.tr/haber/cumhurb EshaHaber.com.tr #Cumhurbaşkanı #Erdoğan #Umman #İran #İsrail

    #cumhurbaskani #erdogan #umman #iran #israil
  19. Jake Spurlock @[email protected] ·

    I Wanted Podcast Transcriptions. iOS 26 Delivered (and Nearly Melted My Phone).

    Testing iOS 26’s on-device speech recognition: faster than realtime, but your phone might disagree

    Apple’s iOS 26 introduced SpeechTranscriber – a promise of on-device, private, offline podcast transcription. No cloud, no subscription, just pure silicon magic. I built it into my RSS reader app. Here’s what actually happened.

    The Setup

    The Good News: It’s Actually Fast

    EpisodeDurationTranscription TimeRealtime FactorWordsWords/secTalk Show #4361h 35m15m 22s6.2x17,30318.8Upgrade #5941h 46m20m 4s5.3x19,97516.6ATP #6681h 54m24m 49s4.6x23,89216.0

    4.6x to 6.2x faster than realtime. Nearly 2-hour podcasts transcribed in under 25 minutes. The Neural Engine absolutely crushes this.

    The Pipeline Breakdown

    The transcription happens in two phases (example from Upgrade #594):

    1. Audio Analysis: 2m 2s
      • Initial pass through the audio file
      • Roughly 1 second of analysis per minute of audio
    2. Results Collection: 18m 0s
      • Iterating through ~1,288 speech segments
      • Each segment yields transcribed text

    The Bad News: Thermal Throttling Is Real

    During my first test, I made a critical mistake: running two transcriptions simultaneously while charging.

    The result? My phone got noticeably hot. Battery optimization warnings appeared. And performance dropped dramatically:

    ConditionRealtime FactorPerformance HitSingle transcription4.6x – 6.2xBaselineTwo parallel transcriptions2.7x46% slower

    The logs showed alternating progress updates as iOS juggled both workloads:

    🎙️ 📝 Progress: 34% - 88 segments   // Transcription A
🎙️ 📝 Progress: 44% - 98 segments   // Transcription B
🎙️ 📝 Progress: 37% - 98 segments   // Transcription A

    The Neural Engine throttles hard when thermals get bad. When I ran a single transcription without charging, the ETA stayed consistent and completed on schedule.

    The Ugly: iOS Kills Background Tasks

    Even with BGTaskScheduler, iOS terminated my background transcription:

    🎙️ Background transcription task triggered by iOS
⏱️ Background transcription task expired (iOS terminated it)

    For long podcasts, you need to keep the app in foreground. iOS’s aggressive app suspension doesn’t play nice with hour-long ML workloads.

    AI Chapter Generation: The Real Win

    Here’s where it gets interesting. Once you have a transcript, generating AI chapters is blazingly fast.

    Note: ATP, Talk Show, and Upgrade already include chapters via ID3 tags – this is an experiment to see what on-device AI can generate. But Planet Money doesn’t have chapters, making it a real use case where AI generation adds genuine value.

    And we’re not alone in this approach. As Mike Hurley and Jason Snell discussed on Upgrade #594, Apple is doing exactly this in iOS 26.2’s Podcasts app:

    “One of the most interesting things to me is the changes in the podcast app in 26.2… AI generated chapters for podcasts that do not support them… They are creating their own chapters based on the topics.”

    Jason nailed the insight: “The transcripts [are] a feature that unlocks a lot of other features, because now they kind of understand the content of the podcast.”

    That’s exactly what we’re doing here – using on-device transcription as a foundation for AI-powered chapter generation:

    EpisodeTranscript SizeChapters GeneratedTimeATP #669143,603 chars (~26,387 words)27 chapters2m 1sTalk Show #436~17,303 words13 chapters1m 40s

    The AI identified topic changes, extracted key phrases for timestamps, and generated descriptive chapter titles – all in under 2 minutes for multi-hour podcasts.

    Sample generated chapters:

    📍 0:00-2:18: Snowfall in Richmond
📍 42:43-49:11: Intel-Apple Chip Collaboration Speculations
📍 62:46-65:00: Executive Transitions at Apple
📍 95:56-105:04: Core Values and Apple's Evolution

    The Code

    Using iOS 26’s SpeechTranscriber is surprisingly clean:

    @available(iOS 26.0, *)
func transcribe(fileURL: URL) async throws -> String {
    let locale = try await findSupportedLocale(preferring: "en")
    let transcriber = SpeechTranscriber(locale: locale, preset: .transcription)
    let analyzer = SpeechAnalyzer(modules: [transcriber])

    let audioFile = try AVAudioFile(forReading: fileURL)
    if let lastSample = try await analyzer.analyzeSequence(from: audioFile) {
        try await analyzer.finalizeAndFinish(through: lastSample)
    }

    var transcription = ""
    for try await result in transcriber.results {
        if result.isFinal {
            transcription += String(result.text.characters) + " "
        }
    }
    return transcription
}

    Fast vs Accurate Mode: A Surprising Finding

    iOS 26 offers two main transcription presets:

    • .transcription – Standard accurate mode
    • .progressiveTranscription – “Fast” mode with progressive results

    I assumed Fast mode would be… faster. The results were mixed.

    EpisodeModeConditionRealtime FactorWords/secTalk Show #436AccurateSolo, cold6.2x18.8Upgrade #594AccurateSolo5.3x16.6ATP #668AccurateSolo4.6x16.0Planet MoneyFastSolo3.8x12.2Planet MoneyAccurateSolo, warm3.5x11.4

    On the same 31-minute episode, Fast mode (3.8x) was slightly faster than Accurate (3.5x). But both were significantly slower than the longer episode tests – likely due to residual heat from previous runs.

    The “progressive” preset appears optimized for live/streaming transcription. For batch processing of pre-recorded files, results are similar when thermals are equivalent.

    Lesson: Don’t assume “fast” means faster for your use case. Profile both.

    Recommendations

    1. Use .transcription for downloaded files – It’s actually faster for batch processing
    2. Don’t charge while transcribing – Thermal throttling is real
    3. One transcription at a time – The Neural Engine doesn’t parallelize well
    4. Keep the app in foreground – iOS will kill background ML tasks
    5. Expect ~5x realtime – About 12-13 minutes per hour of audio under ideal conditions

    The Verdict

    iOS 26’s on-device transcription is genuinely impressive:

    • Privacy: Audio never leaves your device
    • Speed: 5x faster than realtime (when not throttled)
    • Quality: Surprisingly accurate for conversational podcasts
    • Offline: Once the model is downloaded, no internet required

    The main gotchas are thermal management and iOS’s background task limitations. But for a first-generation on-device transcription API? Apple’s Neural Engine delivers.

    Now if you’ll excuse me, I have 26,387 words of ATP to search through.

    Tested on iPhone 17 Pro Max running iOS 26.x. Your mileage may vary on older devices.

    Raw Test Data

    Upgrade #594

    • Audio Duration: 1h 46m 24s (106 min)
    • Audio Analysis Phase: 2m 2s
    • Results Collection Phase: 18m 0s
    • Total Transcription Time: 20m 4s
    • Realtime Factor: 5.3x (faster than audio playback)
    • Words Transcribed: 19,975
    • Processing Rate: 16.6 words/sec
    • Segments Processed: 1,288

    ATP #668

    • Audio Duration: 1h 53m 54s (114 min)
    • Audio Analysis Phase: 2m 20s
    • Results Collection Phase: 22m 28s
    • Total Transcription Time: 24m 49s
    • Realtime Factor: 4.6x (faster than audio playback)
    • Words Transcribed: 23,892
    • Processing Rate: 16.0 words/sec
    • Segments Processed: 1,557

    ATP #669 Chapter Generation

    • Audio Duration: 2h 2m 13s (122 min)
    • Transcription Size: 143,603 characters, ~26,387 words
    • Chapters Generated: 27
    • Total Time: 2m 1s
    • Processing Rate: ~219 words/sec

    Talk Show #436

    • Audio Duration: 1h 35m 52s (95 min)
    • Audio Analysis Phase: 1m 37s
    • Results Collection Phase: 13m 44s
    • Total Transcription Time: 15m 22s
    • Realtime Factor: 6.2x (faster than audio playback) ← Fastest test!
    • Words Transcribed: 17,303
    • Processing Rate: 18.8 words/sec
    • Segments Processed: 971

    Talk Show #436 Chapter Generation

    • Transcription Size: ~17,303 words
    • Chapters Generated: 13
    • Total Time: 1m 40s

    Planet Money – Chicago Parking Meters (Fast Mode)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 3s
    • Results Collection Phase: 7m 5s
    • Total Transcription Time: 8m 9s
    • Realtime Factor: 3.8x
    • Words Transcribed: 5,981
    • Processing Rate: 12.2 words/sec
    • Segments Processed: 472
    • Mode: .progressiveTranscription (Fast)

    Planet Money Chapter Generation (Fast Mode)

    • Transcription Size: ~5,981 words
    • Chapters Generated: 8
    • Total Time: 31.9 sec

    Planet Money – Accurate Mode (Parallel Stress Test)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 9s
    • Results Collection Phase: 10m 8s
    • Total Transcription Time: 11m 19s
    • Realtime Factor: 2.7x ← Severely throttled (ran 2 simultaneous)
    • Words Transcribed: 5,983
    • Processing Rate: 8.8 words/sec
    • Segments Processed: 476
    • Mode: .transcription (Accurate)
    • Note: Ran in parallel with another transcription – 46% performance hit

    Planet Money – Accurate Mode (Solo, Warm Device)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 11s
    • Results Collection Phase: 7m 32s
    • Total Transcription Time: 8m 44s
    • Realtime Factor: 3.5x ← Device still warm from previous tests
    • Words Transcribed: 5,983
    • Processing Rate: 11.4 words/sec
    • Segments Processed: 477
    • Mode: .transcription (Accurate)
    • Note: Slightly slower than Fast mode on same episode (thermal impact)

    Device Observations

    • Thermal: Significant heat when running multiple transcriptions while charging
    • Thermal Carryover: Running tests back-to-back shows degraded performance (6.2x cold → 3.5x warm)
    • Cool-down Recommended: Wait 5-10 minutes between long transcriptions for optimal performance
    • Battery Notifications: Battery optimization warnings triggered during parallel operations
    • Background Tasks: iOS terminated BGTaskScheduler tasks during long transcriptions
    • Beta Warning: Cannot use modules with unallocated locales [en_US (fixed en_US)] – appears in logs but doesn’t block functionality
    #436 #4361h #436AccurateSolo #594 #5941h #594AccurateSolo5 #668 #6681h #668AccurateSolo4 #669 #669143 #AppleIntelligence #iOS26 #NeuralEngine #onDeviceML #podcastTranscription #SpeechRecognition #SpeechTranscriber #Swift
    #4361h #436accuratesolo #5941h #594accuratesolo5 #6681h #668accuratesolo4
  20. Jake Spurlock @[email protected] ·

    I Wanted Podcast Transcriptions. iOS 26 Delivered (and Nearly Melted My Phone).

    Testing iOS 26’s on-device speech recognition: faster than realtime, but your phone might disagree

    Apple’s iOS 26 introduced SpeechTranscriber – a promise of on-device, private, offline podcast transcription. No cloud, no subscription, just pure silicon magic. I built it into my RSS reader app. Here’s what actually happened.

    The Setup

    • Device: iPhone 17 Pro Max (Orange, if you’re curious)
    • iOS Version: 26.2
    • Test Episodes:
      • Talk Show #436 (95 minutes)
      • Upgrade #594 (106 minutes)
      • ATP #668 (114 minutes)
      • ATP #669 (122 minutes)

    The Good News: It’s Actually Fast

    EpisodeDurationTranscription TimeRealtime FactorWordsWords/secTalk Show #4361h 35m15m 22s6.2x17,30318.8Upgrade #5941h 46m20m 4s5.3x19,97516.6ATP #6681h 54m24m 49s4.6x23,89216.0

    4.6x to 6.2x faster than realtime. Nearly 2-hour podcasts transcribed in under 25 minutes. The Neural Engine absolutely crushes this.

    The Pipeline Breakdown

    The transcription happens in two phases (example from Upgrade #594):

    1. Audio Analysis: 2m 2s
      • Initial pass through the audio file
      • Roughly 1 second of analysis per minute of audio
    2. Results Collection: 18m 0s
      • Iterating through ~1,288 speech segments
      • Each segment yields transcribed text

    The Bad News: Thermal Throttling Is Real

    During my first test, I made a critical mistake: running two transcriptions simultaneously while charging.

    The result? My phone got noticeably hot. Battery optimization warnings appeared. And performance dropped dramatically:

    ConditionRealtime FactorPerformance HitSingle transcription4.6x – 6.2xBaselineTwo parallel transcriptions2.7x46% slower

    The logs showed alternating progress updates as iOS juggled both workloads:

    🎙️ 📝 Progress: 34% - 88 segments   // Transcription A🎙️ 📝 Progress: 44% - 98 segments   // Transcription B🎙️ 📝 Progress: 37% - 98 segments   // Transcription A

    The Neural Engine throttles hard when thermals get bad. When I ran a single transcription without charging, the ETA stayed consistent and completed on schedule.

    The Ugly: iOS Kills Background Tasks

    Even with BGTaskScheduler, iOS terminated my background transcription:

    🎙️ Background transcription task triggered by iOS⏱️ Background transcription task expired (iOS terminated it)

    For long podcasts, you need to keep the app in foreground. iOS’s aggressive app suspension doesn’t play nice with hour-long ML workloads.

    AI Chapter Generation: The Real Win

    Here’s where it gets interesting. Once you have a transcript, generating AI chapters is blazingly fast.

    Note: ATP, Talk Show, and Upgrade already include chapters via ID3 tags – this is an experiment to see what on-device AI can generate. But Planet Money doesn’t have chapters, making it a real use case where AI generation adds genuine value.

    And we’re not alone in this approach. As Mike Hurley and Jason Snell discussed on Upgrade #594, Apple is doing exactly this in iOS 26.2’s Podcasts app:

    “One of the most interesting things to me is the changes in the podcast app in 26.2… AI generated chapters for podcasts that do not support them… They are creating their own chapters based on the topics.”

    Jason nailed the insight: “The transcripts [are] a feature that unlocks a lot of other features, because now they kind of understand the content of the podcast.”

    That’s exactly what we’re doing here – using on-device transcription as a foundation for AI-powered chapter generation:

    EpisodeTranscript SizeChapters GeneratedTimeATP #669143,603 chars (~26,387 words)27 chapters2m 1sTalk Show #436~17,303 words13 chapters1m 40s

    The AI identified topic changes, extracted key phrases for timestamps, and generated descriptive chapter titles – all in under 2 minutes for multi-hour podcasts.

    Sample generated chapters:

    📍 0:00-2:18: Snowfall in Richmond📍 42:43-49:11: Intel-Apple Chip Collaboration Speculations📍 62:46-65:00: Executive Transitions at Apple📍 95:56-105:04: Core Values and Apple's Evolution

    The Code

    Using iOS 26’s SpeechTranscriber is surprisingly clean:

    @available(iOS 26.0, *)func transcribe(fileURL: URL) async throws -> String {    let locale = try await findSupportedLocale(preferring: "en")    let transcriber = SpeechTranscriber(locale: locale, preset: .transcription)    let analyzer = SpeechAnalyzer(modules: [transcriber])    let audioFile = try AVAudioFile(forReading: fileURL)    if let lastSample = try await analyzer.analyzeSequence(from: audioFile) {        try await analyzer.finalizeAndFinish(through: lastSample)    }    var transcription = ""    for try await result in transcriber.results {        if result.isFinal {            transcription += String(result.text.characters) + " "        }    }    return transcription}

    Fast vs Accurate Mode: A Surprising Finding

    iOS 26 offers two main transcription presets:

    • .transcription – Standard accurate mode
    • .progressiveTranscription – “Fast” mode with progressive results

    I assumed Fast mode would be… faster. The results were mixed.

    EpisodeModeConditionRealtime FactorWords/secTalk Show #436AccurateSolo, cold6.2x18.8Upgrade #594AccurateSolo5.3x16.6ATP #668AccurateSolo4.6x16.0Planet MoneyFastSolo3.8x12.2Planet MoneyAccurateSolo, warm3.5x11.4

    On the same 31-minute episode, Fast mode (3.8x) was slightly faster than Accurate (3.5x). But both were significantly slower than the longer episode tests – likely due to residual heat from previous runs.

    The “progressive” preset appears optimized for live/streaming transcription. For batch processing of pre-recorded files, results are similar when thermals are equivalent.

    Lesson: Don’t assume “fast” means faster for your use case. Profile both.

    Recommendations

    1. Use .transcription for downloaded files – It’s actually faster for batch processing
    2. Don’t charge while transcribing – Thermal throttling is real
    3. One transcription at a time – The Neural Engine doesn’t parallelize well
    4. Keep the app in foreground – iOS will kill background ML tasks
    5. Expect ~5x realtime – About 12-13 minutes per hour of audio under ideal conditions

    The Verdict

    iOS 26’s on-device transcription is genuinely impressive:

    • Privacy: Audio never leaves your device
    • Speed: 5x faster than realtime (when not throttled)
    • Quality: Surprisingly accurate for conversational podcasts
    • Offline: Once the model is downloaded, no internet required

    The main gotchas are thermal management and iOS’s background task limitations. But for a first-generation on-device transcription API? Apple’s Neural Engine delivers.

    Now if you’ll excuse me, I have 26,387 words of ATP to search through.

    Tested on iPhone 17 Pro Max running iOS 26.x. Your mileage may vary on older devices.

    Raw Test Data

    Upgrade #594

    • Audio Duration: 1h 46m 24s (106 min)
    • Audio Analysis Phase: 2m 2s
    • Results Collection Phase: 18m 0s
    • Total Transcription Time: 20m 4s
    • Realtime Factor: 5.3x (faster than audio playback)
    • Words Transcribed: 19,975
    • Processing Rate: 16.6 words/sec
    • Segments Processed: 1,288

    ATP #668

    • Audio Duration: 1h 53m 54s (114 min)
    • Audio Analysis Phase: 2m 20s
    • Results Collection Phase: 22m 28s
    • Total Transcription Time: 24m 49s
    • Realtime Factor: 4.6x (faster than audio playback)
    • Words Transcribed: 23,892
    • Processing Rate: 16.0 words/sec
    • Segments Processed: 1,557

    ATP #669 Chapter Generation

    • Audio Duration: 2h 2m 13s (122 min)
    • Transcription Size: 143,603 characters, ~26,387 words
    • Chapters Generated: 27
    • Total Time: 2m 1s
    • Processing Rate: ~219 words/sec

    Talk Show #436

    • Audio Duration: 1h 35m 52s (95 min)
    • Audio Analysis Phase: 1m 37s
    • Results Collection Phase: 13m 44s
    • Total Transcription Time: 15m 22s
    • Realtime Factor: 6.2x (faster than audio playback) ← Fastest test!
    • Words Transcribed: 17,303
    • Processing Rate: 18.8 words/sec
    • Segments Processed: 971

    Talk Show #436 Chapter Generation

    • Transcription Size: ~17,303 words
    • Chapters Generated: 13
    • Total Time: 1m 40s

    Planet Money – Chicago Parking Meters (Fast Mode)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 3s
    • Results Collection Phase: 7m 5s
    • Total Transcription Time: 8m 9s
    • Realtime Factor: 3.8x
    • Words Transcribed: 5,981
    • Processing Rate: 12.2 words/sec
    • Segments Processed: 472
    • Mode: .progressiveTranscription (Fast)

    Planet Money Chapter Generation (Fast Mode)

    • Transcription Size: ~5,981 words
    • Chapters Generated: 8
    • Total Time: 31.9 sec

    Planet Money – Accurate Mode (Parallel Stress Test)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 9s
    • Results Collection Phase: 10m 8s
    • Total Transcription Time: 11m 19s
    • Realtime Factor: 2.7x ← Severely throttled (ran 2 simultaneous)
    • Words Transcribed: 5,983
    • Processing Rate: 8.8 words/sec
    • Segments Processed: 476
    • Mode: .transcription (Accurate)
    • Note: Ran in parallel with another transcription – 46% performance hit

    Planet Money – Accurate Mode (Solo, Warm Device)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 11s
    • Results Collection Phase: 7m 32s
    • Total Transcription Time: 8m 44s
    • Realtime Factor: 3.5x ← Device still warm from previous tests
    • Words Transcribed: 5,983
    • Processing Rate: 11.4 words/sec
    • Segments Processed: 477
    • Mode: .transcription (Accurate)
    • Note: Slightly slower than Fast mode on same episode (thermal impact)

    Device Observations

    • Thermal: Significant heat when running multiple transcriptions while charging
    • Thermal Carryover: Running tests back-to-back shows degraded performance (6.2x cold → 3.5x warm)
    • Cool-down Recommended: Wait 5-10 minutes between long transcriptions for optimal performance
    • Battery Notifications: Battery optimization warnings triggered during parallel operations
    • Background Tasks: iOS terminated BGTaskScheduler tasks during long transcriptions
    • Beta Warning: Cannot use modules with unallocated locales [en_US (fixed en_US)] – appears in logs but doesn’t block functionality

    #436 #4361h #436AccurateSolo #594 #5941h #594AccurateSolo5 #668 #6681h #668AccurateSolo4 #669 #669143 #AppleIntelligence #iOS26 #NeuralEngine #onDeviceML #podcastTranscription #SpeechRecognition #SpeechTranscriber #Swift

    #4361h #436accuratesolo #5941h #594accuratesolo5 #6681h #668accuratesolo4
  21. Jake Spurlock @[email protected] ·

    I Wanted Podcast Transcriptions. iOS 26 Delivered (and Nearly Melted My Phone).

    Testing iOS 26’s on-device speech recognition: faster than realtime, but your phone might disagree

    Apple’s iOS 26 introduced SpeechTranscriber – a promise of on-device, private, offline podcast transcription. No cloud, no subscription, just pure silicon magic. I built it into my RSS reader app. Here’s what actually happened.

    The Setup

    • Device: iPhone 17 Pro Max (Orange, if you’re curious)
    • iOS Version: 26.2
    • Test Episodes:
      • Talk Show #436 (95 minutes)
      • Upgrade #594 (106 minutes)
      • ATP #668 (114 minutes)
      • ATP #669 (122 minutes)

    The Good News: It’s Actually Fast

    EpisodeDurationTranscription TimeRealtime FactorWordsWords/secTalk Show #4361h 35m15m 22s6.2x17,30318.8Upgrade #5941h 46m20m 4s5.3x19,97516.6ATP #6681h 54m24m 49s4.6x23,89216.0

    4.6x to 6.2x faster than realtime. Nearly 2-hour podcasts transcribed in under 25 minutes. The Neural Engine absolutely crushes this.

    The Pipeline Breakdown

    The transcription happens in two phases (example from Upgrade #594):

    1. Audio Analysis: 2m 2s
      • Initial pass through the audio file
      • Roughly 1 second of analysis per minute of audio
    2. Results Collection: 18m 0s
      • Iterating through ~1,288 speech segments
      • Each segment yields transcribed text

    The Bad News: Thermal Throttling Is Real

    During my first test, I made a critical mistake: running two transcriptions simultaneously while charging.

    The result? My phone got noticeably hot. Battery optimization warnings appeared. And performance dropped dramatically:

    ConditionRealtime FactorPerformance HitSingle transcription4.6x – 6.2xBaselineTwo parallel transcriptions2.7x46% slower

    The logs showed alternating progress updates as iOS juggled both workloads:

    🎙️ 📝 Progress: 34% - 88 segments   // Transcription A🎙️ 📝 Progress: 44% - 98 segments   // Transcription B🎙️ 📝 Progress: 37% - 98 segments   // Transcription A

    The Neural Engine throttles hard when thermals get bad. When I ran a single transcription without charging, the ETA stayed consistent and completed on schedule.

    The Ugly: iOS Kills Background Tasks

    Even with BGTaskScheduler, iOS terminated my background transcription:

    🎙️ Background transcription task triggered by iOS⏱️ Background transcription task expired (iOS terminated it)

    For long podcasts, you need to keep the app in foreground. iOS’s aggressive app suspension doesn’t play nice with hour-long ML workloads.

    AI Chapter Generation: The Real Win

    Here’s where it gets interesting. Once you have a transcript, generating AI chapters is blazingly fast.

    Note: ATP, Talk Show, and Upgrade already include chapters via ID3 tags – this is an experiment to see what on-device AI can generate. But Planet Money doesn’t have chapters, making it a real use case where AI generation adds genuine value.

    And we’re not alone in this approach. As Mike Hurley and Jason Snell discussed on Upgrade #594, Apple is doing exactly this in iOS 26.2’s Podcasts app:

    “One of the most interesting things to me is the changes in the podcast app in 26.2… AI generated chapters for podcasts that do not support them… They are creating their own chapters based on the topics.”

    Jason nailed the insight: “The transcripts [are] a feature that unlocks a lot of other features, because now they kind of understand the content of the podcast.”

    That’s exactly what we’re doing here – using on-device transcription as a foundation for AI-powered chapter generation:

    EpisodeTranscript SizeChapters GeneratedTimeATP #669143,603 chars (~26,387 words)27 chapters2m 1sTalk Show #436~17,303 words13 chapters1m 40s

    The AI identified topic changes, extracted key phrases for timestamps, and generated descriptive chapter titles – all in under 2 minutes for multi-hour podcasts.

    Sample generated chapters:

    📍 0:00-2:18: Snowfall in Richmond📍 42:43-49:11: Intel-Apple Chip Collaboration Speculations📍 62:46-65:00: Executive Transitions at Apple📍 95:56-105:04: Core Values and Apple's Evolution

    The Code

    Using iOS 26’s SpeechTranscriber is surprisingly clean:

    @available(iOS 26.0, *)func transcribe(fileURL: URL) async throws -> String {    let locale = try await findSupportedLocale(preferring: "en")    let transcriber = SpeechTranscriber(locale: locale, preset: .transcription)    let analyzer = SpeechAnalyzer(modules: [transcriber])    let audioFile = try AVAudioFile(forReading: fileURL)    if let lastSample = try await analyzer.analyzeSequence(from: audioFile) {        try await analyzer.finalizeAndFinish(through: lastSample)    }    var transcription = ""    for try await result in transcriber.results {        if result.isFinal {            transcription += String(result.text.characters) + " "        }    }    return transcription}

    Fast vs Accurate Mode: A Surprising Finding

    iOS 26 offers two main transcription presets:

    • .transcription – Standard accurate mode
    • .progressiveTranscription – “Fast” mode with progressive results

    I assumed Fast mode would be… faster. The results were mixed.

    EpisodeModeConditionRealtime FactorWords/secTalk Show #436AccurateSolo, cold6.2x18.8Upgrade #594AccurateSolo5.3x16.6ATP #668AccurateSolo4.6x16.0Planet MoneyFastSolo3.8x12.2Planet MoneyAccurateSolo, warm3.5x11.4

    On the same 31-minute episode, Fast mode (3.8x) was slightly faster than Accurate (3.5x). But both were significantly slower than the longer episode tests – likely due to residual heat from previous runs.

    The “progressive” preset appears optimized for live/streaming transcription. For batch processing of pre-recorded files, results are similar when thermals are equivalent.

    Lesson: Don’t assume “fast” means faster for your use case. Profile both.

    Recommendations

    1. Use .transcription for downloaded files – It’s actually faster for batch processing
    2. Don’t charge while transcribing – Thermal throttling is real
    3. One transcription at a time – The Neural Engine doesn’t parallelize well
    4. Keep the app in foreground – iOS will kill background ML tasks
    5. Expect ~5x realtime – About 12-13 minutes per hour of audio under ideal conditions

    The Verdict

    iOS 26’s on-device transcription is genuinely impressive:

    • Privacy: Audio never leaves your device
    • Speed: 5x faster than realtime (when not throttled)
    • Quality: Surprisingly accurate for conversational podcasts
    • Offline: Once the model is downloaded, no internet required

    The main gotchas are thermal management and iOS’s background task limitations. But for a first-generation on-device transcription API? Apple’s Neural Engine delivers.

    Now if you’ll excuse me, I have 26,387 words of ATP to search through.

    Tested on iPhone 17 Pro Max running iOS 26.x. Your mileage may vary on older devices.

    Raw Test Data

    Upgrade #594

    • Audio Duration: 1h 46m 24s (106 min)
    • Audio Analysis Phase: 2m 2s
    • Results Collection Phase: 18m 0s
    • Total Transcription Time: 20m 4s
    • Realtime Factor: 5.3x (faster than audio playback)
    • Words Transcribed: 19,975
    • Processing Rate: 16.6 words/sec
    • Segments Processed: 1,288

    ATP #668

    • Audio Duration: 1h 53m 54s (114 min)
    • Audio Analysis Phase: 2m 20s
    • Results Collection Phase: 22m 28s
    • Total Transcription Time: 24m 49s
    • Realtime Factor: 4.6x (faster than audio playback)
    • Words Transcribed: 23,892
    • Processing Rate: 16.0 words/sec
    • Segments Processed: 1,557

    ATP #669 Chapter Generation

    • Audio Duration: 2h 2m 13s (122 min)
    • Transcription Size: 143,603 characters, ~26,387 words
    • Chapters Generated: 27
    • Total Time: 2m 1s
    • Processing Rate: ~219 words/sec

    Talk Show #436

    • Audio Duration: 1h 35m 52s (95 min)
    • Audio Analysis Phase: 1m 37s
    • Results Collection Phase: 13m 44s
    • Total Transcription Time: 15m 22s
    • Realtime Factor: 6.2x (faster than audio playback) ← Fastest test!
    • Words Transcribed: 17,303
    • Processing Rate: 18.8 words/sec
    • Segments Processed: 971

    Talk Show #436 Chapter Generation

    • Transcription Size: ~17,303 words
    • Chapters Generated: 13
    • Total Time: 1m 40s

    Planet Money – Chicago Parking Meters (Fast Mode)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 3s
    • Results Collection Phase: 7m 5s
    • Total Transcription Time: 8m 9s
    • Realtime Factor: 3.8x
    • Words Transcribed: 5,981
    • Processing Rate: 12.2 words/sec
    • Segments Processed: 472
    • Mode: .progressiveTranscription (Fast)

    Planet Money Chapter Generation (Fast Mode)

    • Transcription Size: ~5,981 words
    • Chapters Generated: 8
    • Total Time: 31.9 sec

    Planet Money – Accurate Mode (Parallel Stress Test)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 9s
    • Results Collection Phase: 10m 8s
    • Total Transcription Time: 11m 19s
    • Realtime Factor: 2.7x ← Severely throttled (ran 2 simultaneous)
    • Words Transcribed: 5,983
    • Processing Rate: 8.8 words/sec
    • Segments Processed: 476
    • Mode: .transcription (Accurate)
    • Note: Ran in parallel with another transcription – 46% performance hit

    Planet Money – Accurate Mode (Solo, Warm Device)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 11s
    • Results Collection Phase: 7m 32s
    • Total Transcription Time: 8m 44s
    • Realtime Factor: 3.5x ← Device still warm from previous tests
    • Words Transcribed: 5,983
    • Processing Rate: 11.4 words/sec
    • Segments Processed: 477
    • Mode: .transcription (Accurate)
    • Note: Slightly slower than Fast mode on same episode (thermal impact)

    Device Observations

    • Thermal: Significant heat when running multiple transcriptions while charging
    • Thermal Carryover: Running tests back-to-back shows degraded performance (6.2x cold → 3.5x warm)
    • Cool-down Recommended: Wait 5-10 minutes between long transcriptions for optimal performance
    • Battery Notifications: Battery optimization warnings triggered during parallel operations
    • Background Tasks: iOS terminated BGTaskScheduler tasks during long transcriptions
    • Beta Warning: Cannot use modules with unallocated locales [en_US (fixed en_US)] – appears in logs but doesn’t block functionality

    #436 #4361h #436AccurateSolo #594 #5941h #594AccurateSolo5 #668 #6681h #668AccurateSolo4 #669 #669143 #AppleIntelligence #iOS26 #NeuralEngine #onDeviceML #podcastTranscription #SpeechRecognition #SpeechTranscriber #Swift

    #swift #speechtranscriber #speechrecognition #podcasttranscription #ondeviceml #neuralengine
  22. Jake Spurlock @[email protected] ·

    I Wanted Podcast Transcriptions. iOS 26 Delivered (and Nearly Melted My Phone).

    Testing iOS 26’s on-device speech recognition: faster than realtime, but your phone might disagree

    Apple’s iOS 26 introduced SpeechTranscriber – a promise of on-device, private, offline podcast transcription. No cloud, no subscription, just pure silicon magic. I built it into my RSS reader app. Here’s what actually happened.

    The Setup

    • Device: iPhone 17 Pro Max (Orange, if you’re curious)
    • iOS Version: 26.2
    • Test Episodes:
      • Talk Show #436 (95 minutes)
      • Upgrade #594 (106 minutes)
      • ATP #668 (114 minutes)
      • ATP #669 (122 minutes)

    The Good News: It’s Actually Fast

    EpisodeDurationTranscription TimeRealtime FactorWordsWords/secTalk Show #4361h 35m15m 22s6.2x17,30318.8Upgrade #5941h 46m20m 4s5.3x19,97516.6ATP #6681h 54m24m 49s4.6x23,89216.0

    4.6x to 6.2x faster than realtime. Nearly 2-hour podcasts transcribed in under 25 minutes. The Neural Engine absolutely crushes this.

    The Pipeline Breakdown

    The transcription happens in two phases (example from Upgrade #594):

    1. Audio Analysis: 2m 2s
      • Initial pass through the audio file
      • Roughly 1 second of analysis per minute of audio
    2. Results Collection: 18m 0s
      • Iterating through ~1,288 speech segments
      • Each segment yields transcribed text

    The Bad News: Thermal Throttling Is Real

    During my first test, I made a critical mistake: running two transcriptions simultaneously while charging.

    The result? My phone got noticeably hot. Battery optimization warnings appeared. And performance dropped dramatically:

    ConditionRealtime FactorPerformance HitSingle transcription4.6x – 6.2xBaselineTwo parallel transcriptions2.7x46% slower

    The logs showed alternating progress updates as iOS juggled both workloads:

    🎙️ 📝 Progress: 34% - 88 segments   // Transcription A🎙️ 📝 Progress: 44% - 98 segments   // Transcription B🎙️ 📝 Progress: 37% - 98 segments   // Transcription A

    The Neural Engine throttles hard when thermals get bad. When I ran a single transcription without charging, the ETA stayed consistent and completed on schedule.

    The Ugly: iOS Kills Background Tasks

    Even with BGTaskScheduler, iOS terminated my background transcription:

    🎙️ Background transcription task triggered by iOS⏱️ Background transcription task expired (iOS terminated it)

    For long podcasts, you need to keep the app in foreground. iOS’s aggressive app suspension doesn’t play nice with hour-long ML workloads.

    AI Chapter Generation: The Real Win

    Here’s where it gets interesting. Once you have a transcript, generating AI chapters is blazingly fast.

    Note: ATP, Talk Show, and Upgrade already include chapters via ID3 tags – this is an experiment to see what on-device AI can generate. But Planet Money doesn’t have chapters, making it a real use case where AI generation adds genuine value.

    And we’re not alone in this approach. As Mike Hurley and Jason Snell discussed on Upgrade #594, Apple is doing exactly this in iOS 26.2’s Podcasts app:

    “One of the most interesting things to me is the changes in the podcast app in 26.2… AI generated chapters for podcasts that do not support them… They are creating their own chapters based on the topics.”

    Jason nailed the insight: “The transcripts [are] a feature that unlocks a lot of other features, because now they kind of understand the content of the podcast.”

    That’s exactly what we’re doing here – using on-device transcription as a foundation for AI-powered chapter generation:

    EpisodeTranscript SizeChapters GeneratedTimeATP #669143,603 chars (~26,387 words)27 chapters2m 1sTalk Show #436~17,303 words13 chapters1m 40s

    The AI identified topic changes, extracted key phrases for timestamps, and generated descriptive chapter titles – all in under 2 minutes for multi-hour podcasts.

    Sample generated chapters:

    📍 0:00-2:18: Snowfall in Richmond📍 42:43-49:11: Intel-Apple Chip Collaboration Speculations📍 62:46-65:00: Executive Transitions at Apple📍 95:56-105:04: Core Values and Apple's Evolution

    The Code

    Using iOS 26’s SpeechTranscriber is surprisingly clean:

    @available(iOS 26.0, *)func transcribe(fileURL: URL) async throws -> String {    let locale = try await findSupportedLocale(preferring: "en")    let transcriber = SpeechTranscriber(locale: locale, preset: .transcription)    let analyzer = SpeechAnalyzer(modules: [transcriber])    let audioFile = try AVAudioFile(forReading: fileURL)    if let lastSample = try await analyzer.analyzeSequence(from: audioFile) {        try await analyzer.finalizeAndFinish(through: lastSample)    }    var transcription = ""    for try await result in transcriber.results {        if result.isFinal {            transcription += String(result.text.characters) + " "        }    }    return transcription}

    Fast vs Accurate Mode: A Surprising Finding

    iOS 26 offers two main transcription presets:

    • .transcription – Standard accurate mode
    • .progressiveTranscription – “Fast” mode with progressive results

    I assumed Fast mode would be… faster. The results were mixed.

    EpisodeModeConditionRealtime FactorWords/secTalk Show #436AccurateSolo, cold6.2x18.8Upgrade #594AccurateSolo5.3x16.6ATP #668AccurateSolo4.6x16.0Planet MoneyFastSolo3.8x12.2Planet MoneyAccurateSolo, warm3.5x11.4

    On the same 31-minute episode, Fast mode (3.8x) was slightly faster than Accurate (3.5x). But both were significantly slower than the longer episode tests – likely due to residual heat from previous runs.

    The “progressive” preset appears optimized for live/streaming transcription. For batch processing of pre-recorded files, results are similar when thermals are equivalent.

    Lesson: Don’t assume “fast” means faster for your use case. Profile both.

    Recommendations

    1. Use .transcription for downloaded files – It’s actually faster for batch processing
    2. Don’t charge while transcribing – Thermal throttling is real
    3. One transcription at a time – The Neural Engine doesn’t parallelize well
    4. Keep the app in foreground – iOS will kill background ML tasks
    5. Expect ~5x realtime – About 12-13 minutes per hour of audio under ideal conditions

    The Verdict

    iOS 26’s on-device transcription is genuinely impressive:

    • Privacy: Audio never leaves your device
    • Speed: 5x faster than realtime (when not throttled)
    • Quality: Surprisingly accurate for conversational podcasts
    • Offline: Once the model is downloaded, no internet required

    The main gotchas are thermal management and iOS’s background task limitations. But for a first-generation on-device transcription API? Apple’s Neural Engine delivers.

    Now if you’ll excuse me, I have 26,387 words of ATP to search through.

    Tested on iPhone 17 Pro Max running iOS 26.x. Your mileage may vary on older devices.

    Raw Test Data

    Upgrade #594

    • Audio Duration: 1h 46m 24s (106 min)
    • Audio Analysis Phase: 2m 2s
    • Results Collection Phase: 18m 0s
    • Total Transcription Time: 20m 4s
    • Realtime Factor: 5.3x (faster than audio playback)
    • Words Transcribed: 19,975
    • Processing Rate: 16.6 words/sec
    • Segments Processed: 1,288

    ATP #668

    • Audio Duration: 1h 53m 54s (114 min)
    • Audio Analysis Phase: 2m 20s
    • Results Collection Phase: 22m 28s
    • Total Transcription Time: 24m 49s
    • Realtime Factor: 4.6x (faster than audio playback)
    • Words Transcribed: 23,892
    • Processing Rate: 16.0 words/sec
    • Segments Processed: 1,557

    ATP #669 Chapter Generation

    • Audio Duration: 2h 2m 13s (122 min)
    • Transcription Size: 143,603 characters, ~26,387 words
    • Chapters Generated: 27
    • Total Time: 2m 1s
    • Processing Rate: ~219 words/sec

    Talk Show #436

    • Audio Duration: 1h 35m 52s (95 min)
    • Audio Analysis Phase: 1m 37s
    • Results Collection Phase: 13m 44s
    • Total Transcription Time: 15m 22s
    • Realtime Factor: 6.2x (faster than audio playback) ← Fastest test!
    • Words Transcribed: 17,303
    • Processing Rate: 18.8 words/sec
    • Segments Processed: 971

    Talk Show #436 Chapter Generation

    • Transcription Size: ~17,303 words
    • Chapters Generated: 13
    • Total Time: 1m 40s

    Planet Money – Chicago Parking Meters (Fast Mode)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 3s
    • Results Collection Phase: 7m 5s
    • Total Transcription Time: 8m 9s
    • Realtime Factor: 3.8x
    • Words Transcribed: 5,981
    • Processing Rate: 12.2 words/sec
    • Segments Processed: 472
    • Mode: .progressiveTranscription (Fast)

    Planet Money Chapter Generation (Fast Mode)

    • Transcription Size: ~5,981 words
    • Chapters Generated: 8
    • Total Time: 31.9 sec

    Planet Money – Accurate Mode (Parallel Stress Test)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 9s
    • Results Collection Phase: 10m 8s
    • Total Transcription Time: 11m 19s
    • Realtime Factor: 2.7x ← Severely throttled (ran 2 simultaneous)
    • Words Transcribed: 5,983
    • Processing Rate: 8.8 words/sec
    • Segments Processed: 476
    • Mode: .transcription (Accurate)
    • Note: Ran in parallel with another transcription – 46% performance hit

    Planet Money – Accurate Mode (Solo, Warm Device)

    • Audio Duration: 30m 56s (31 min)
    • Audio Analysis Phase: 1m 11s
    • Results Collection Phase: 7m 32s
    • Total Transcription Time: 8m 44s
    • Realtime Factor: 3.5x ← Device still warm from previous tests
    • Words Transcribed: 5,983
    • Processing Rate: 11.4 words/sec
    • Segments Processed: 477
    • Mode: .transcription (Accurate)
    • Note: Slightly slower than Fast mode on same episode (thermal impact)

    Device Observations

    • Thermal: Significant heat when running multiple transcriptions while charging
    • Thermal Carryover: Running tests back-to-back shows degraded performance (6.2x cold → 3.5x warm)
    • Cool-down Recommended: Wait 5-10 minutes between long transcriptions for optimal performance
    • Battery Notifications: Battery optimization warnings triggered during parallel operations
    • Background Tasks: iOS terminated BGTaskScheduler tasks during long transcriptions
    • Beta Warning: Cannot use modules with unallocated locales [en_US (fixed en_US)] – appears in logs but doesn’t block functionality

    #436 #4361h #436AccurateSolo #594 #5941h #594AccurateSolo5 #668 #6681h #668AccurateSolo4 #669 #669143 #AppleIntelligence #iOS26 #NeuralEngine #onDeviceML #podcastTranscription #SpeechRecognition #SpeechTranscriber #Swift

    #4361h #436accuratesolo #5941h #594accuratesolo5 #6681h #668accuratesolo4
  23. Karolis Žukauskas­🤖 @[email protected] ·
    Seimo narys, per anksti susenęs jaunasis socialdemokratas Ruslanas Baranovas imasi viešai ginti finansinį sukčių (be jokių “galimai” - sukčius ir taškas), Nemuno aušros gaujos vieną iš vadeivų Puchovičių ir gaujos vadą antisemitą Žemaitaitį. Nes matai, pirmasis viešai padejavo, kad nebegali net į statybas nuvažiuoti darbo Seime metu, jaučiasi persekiojamas, ir kad būtų įtikinamiau - į dejones sugalvojo įvelti šeimos kortą.
    Mielas spaliuk Ruslanai, toks bandymas viešai ginti nusikaltėlius, kurie, deja, laimingai sukritusių aplinkybių dėka tapo Seimo nariais, yra naivumo ir durnumo viešas demonstravimas. Manau, reikia sudėti keletą taškų ant i.
    Skųstis viešumu būnant Seimo nariu - kvaila. Kiekvienas žinojote, kur einate. Patys pasirinkote dalies savo privataus gyvenimo atsisakymą vardan šių pareigų. Bet jauskite skirtumą - yra daugybė Seimo narių, kurie nėra nieko prisidirbę, arba kurie net nėra įdomūs rinkėjams. Nemuno aušra pati pasirinko tokį kelią - destrukcinį, melo, apgaulės ir įžūlių varkių schemų. Tokių įžūlių, kad su kuo bekalbėčiau, visi sako tą patį - šie politikos marozai yra tokie, kokių iki šiol nesame turėję. Darbo partijos schemos buvo subtilesnės, ir tarp jų buvo vienas kitas narys, kuris pagalvodavo apie valstybės reikalus. Pakso šutvėje irgi ne visi buvo susitepę. Kokie nors violetiniai, atėję į Seimą iš Garliavos, gal buvo išprotėję politiškai, bet neprisivogė ir nespjaudė rinkėjams į akis.
    Tuo tarpu Nemuno aušra - ir sakau tai labai atsakingai - yra totalus finansinės gaujos darinys, kuriam nerūpi nė vienas valstybės klausimas. Aušriečiai į Seimą atėjo pasitvarkyti savo reikalų - kas tarnauti rimtesniems rėmėjams politikos pagalba išsprendžiant jų makro lygio problemas, kertančias valstybines sienas. Kas panoro daugiau užsidirbti sau, nes Seimo nario alga yra didesnė nei kokio kūno kultūros mokytojo, automobilių ar kebabų pardavėjo. Kas pamanė, kad pasitvarkys savo verslo reikalus - statybų leidimais, ir pan.
    Nemuno aušros ministerijos tapo paslaugų pardavėjomis - yra verslo sritys, kurioms aktualu tam tikras reguliavimas, ir jie žino, kad susimokėjus Nemuno aušros veikėjams šias problemas galima išsispręsti. Jų dėka turime tokį politinės prostitucijos kioskelį, kur per langelį parduodamos visos paslaugos - svarbu mokėkit pinigus. Schemos, kaip su jais atsiskaitoma, dar tik pradeda lįsti į paviršių - turiu daugiau informacijos, deja, ne visa ja galiu dalintis viešai, nes paaiškėtų šaltiniai, o jų neišduosiu. Bet net ir tai, kiek iki šiol informacijos paskelbiau, jokiam blaivią galvą turinčiam asmeniui neleistų pulti ir viešai ginti Nemuno aušros veikėjų. Bet štai jaunasis spaliukas pasirenka krūtine užstoti finansinį machinatorių.
    Kas skauda ir ko čia bijot? Kad atostogų metu alaus išgersit ir kažkas jus nufotografuos? Visų pirma, gerkit nors ir kibirais - nei man, nei kitiems neįdomu. Bet jei gersit taip, kaip Egipte su rusais gėrė ir apie trąšas garsiai kalbėjo Valius Ąžuolas - na tai kaltinkit tik save, jei jau pardavinėjat valstybę, tai sugebėkit tai daryti tyliai. Antra, Seime yra 141 narys. Aš aprašiau kiek? Tris keturis. Ir jau pradedat absoliutinti prisidirbusiųjų problemą? Bijot, kad ir jūsų darbeliai išlįs? Bijot atostogauti ar išeiti Vilniuje į barus? Neprisidirbę išeina, atostogauja ir nebijo - kodėl kitiems taip sunku? Aš irgi nesidžiaugu dėl visų šitų nesąmonių viešinimo atsisakęs dalies savo privatumo - mane visokios janutienės turguose stebi ir kuria mitus, Žemaitaičio gerbėjai kasdien keiksmais užverčia, o keliaudamas per Lietuvą vasarą visokiose Žemaitaičio prerijose turbūt dar ir akis keiksmų susilauksiu. Bet aš neverkiu dėl to - pats pasirinkau tokį kelią, nes užkniso jūsų geltonas lietutis mums į akis.
    Neverkit ir jūs - savo noru tapot Seimo nariais, o kartelė Seimo nariams yra labai žema - iš esmės tik trys reikalavimai iš piliečių: būkite skaidrūs, nevokite ir atlikite savo pareigas. Nežinau NĖ VIENO aušriečio Seime, kuris atitiktų bent VIENĄ iš šių trijų reikalavimų. Norit, kad tauta tylėtų visą šitą įžūlią mėšlo kvapo politinę tapybą matydami? Verda visiems kraujas, nes tokios įžūliai meluojančios ir niekšiškos koalicijos nesame turėję. Jei socdemai savo kūnu nori pridengti aušriečių gaują, tai patys kalti - nesiskųskite, jei strėlės lėks ir į jus.
    Normali valdžia pasidžiaugtų aktyviais piliečiais. Bet jums, kaip homo sovieticus mumijoms, aktyvūs piliečiai tik trukdo. Nes visi turit savo varkių - kas pasivogti lėšų kurui, kas įstatymą draugeliams prastumti, kas baltarusiams atidirbti, kas LRT užvaldyti. Buki jūs esat, ir dėl to pykstat ant piliečių, kurie išdrįsta tai pasakyti. Socdemai turėjo ir turi šimtus progų išspirti vagis iš koalicijos, bet tuo nesinaudoja, nusprendė degti kartu - tai ir dekit, bet dekit tyliai, nelodami ant savo piliečių.
    Nesusireikšminkit - sekti jūsų atostogas ar pietums siurbiamą sriubą yra nė kiek neįdomu, ir niekas to nedaro. Bet kai Seimo nariai į akis meluoja rinkėjams, o už akių su šešėlinio pasaulio veikėjais derina reikaliukus, arba savo pareigas iškeičia į statybų priežiūrą - turėkit gėdos neverkti, kad žmonės į jus atkreipia dėmesį, fiksuoja ir viešina. Ir aš tikiuosi, kad ši nauja tradicija liks ilgam - gal tada į politiką tie visi požemio gyventojai nebedrįs lįsti, nes saulės šviesos jie bijo labiau nei teisėsaugos.
    Jei šis Baranovo ašarų įrašas skirtas man (o taip juk ir yra) - galiu nuraminti, pats Ruslanas Baranovas man yra absoliučiai neįdomus - nebent paaiškės, kad vagia iš valstybės kibirais, kaip aušriečiai, tada mielai paviešinsiu, ką sužinosiu.
    Eikit dirbt nežliumbę, politiniai kurmiai.

    ⏩
    #ruporas #feisbukasmastodone
  24. Gytis Repečka @[email protected] ·

    Labai nemėgstu samprotavimų, kad esą Europa technologiškai atsilieka nuo JAV. Melas ir paistalai! :neocat_foxmask:

    „Verslo žinios“ kalbina JAV kontorą „Shift4“, gaminančią mokėjimų sprendimus. Jos įkūrėjas milijardierius Jared Isaacman didžiuojasi užuot mokęsis universitete, pasukęs tiesiai į verslą, ir yra tiek apsėstas kosmoso reikalais, kad prie Trumpo net tapo NASA vadovu. Į panašią dūdą pučia dabartinis įmonės CEO Taylor Lauber:

    Jis su „Shift4“ įkūrėju „tėvų rūsyje“ dirbo dar paauglystėje ir, jo paties žodžiais, buvo „tuo idiotu, kuris nusprendė eiti mokytis į koledžą“. Taip T. Lauberis referuoja į jo ir J. Isaacmano išsiskyrusius kelius bei „Shift4“ įkūrėjo sprendimą nesimokyti universitete ir vietoj to nuo pat paauglystės rinktis antreprenerio kelią.

    Kaip ir galima tikėtis, pažeriama ir daugiau išminties perlų:

    Tokio tipo integruoti mokėjimų sprendimai verslui daug labiau paplitę JAV nei Europoje. T. Lauberis tai aiškina tuo, jog Senajame žemyne istoriškai prioritetas buvo teikiamas vartotojo patirčiai ir saugumui, o ne pardavėjo poreikiams. Nors dabar taip nebėra ir integruoti mokėjimų sprendimai yra saugūs, Europa į šį traukinį vėluoja.
    ...
    Jis teigia, kad įprastiniai atsiskaitymo terminalai, kokius naudojame ir Lietuvoje, yra saugūs ir patrauklūs pirkėjams, bet galų gale „tai tėra banko terminalas“.
    „Jis nėra prijungtas prie kitų dalykų, kurių gali reikėti pardavėjui jo versle. Pavyzdžiui, lojalumo programų ar paskatų tam, kad sugrįžtum pas jį pirkti ateityje. Arba, pavyzdžiui, sistema, kuri naudojama jūsų mėsainio ar alaus užsakymui užfiksuoti bei jam nusiųsti į virtuvę, – ji veikia visiškai atskirai nuo terminalo, kuris jums paduodamas, sistemos“, – aiškina T. Lauberis.

    Iššifruokim ką direktorius vadina „pažanga“: pirmiausia, ne kliento patogumą ir saugumą, o verslo interesus. Būtent dėmesys kliento saugumui ir privatumui yra inovacija, o ne siautėjantys duomenų brokeriai ir nesustojantis kiekvieno žingsnio sekimas, neva pateisinantis agresyvų kišimąsi į žmonių privatų gyvenimą dėl kažin kokios menkutės lojalumo nuolaidos. Visiškas akių dūmimas :neocat_thumbsdown:

    JAV prireikė labai daug laiko išlipti iš magnetinių mokėjimo kortelių - būtent JAV dešimtmečiais atsilieka nuo Europos. Bet, žinoma, milijardieriai norėtų, kad visi manytų kitaip :neocat_glare:

    #verslas #mokėjimai #Lietuvoje #aktualijos #shift4

    #verslas #mokejimai #lietuvoje #aktualijos #shift4
  25. Gytis Repečka @[email protected] ·

    Labai nemėgstu samprotavimų, kad esą Europa technologiškai atsilieka nuo JAV. Melas ir paistalai! :neocat_foxmask:

    „Verslo žinios“ kalbina JAV kontorą „Shift4“, gaminančią mokėjimų sprendimus. Jos įkūrėjas milijardierius Jared Isaacman didžiuojasi užuot mokęsis universitete, pasukęs tiesiai į verslą, ir yra tiek apsėstas kosmoso reikalais, kad prie Trumpo net tapo NASA vadovu. Į panašią dūdą pučia dabartinis įmonės CEO Taylor Lauber:

    Jis su „Shift4“ įkūrėju „tėvų rūsyje“ dirbo dar paauglystėje ir, jo paties žodžiais, buvo „tuo idiotu, kuris nusprendė eiti mokytis į koledžą“. Taip T. Lauberis referuoja į jo ir J. Isaacmano išsiskyrusius kelius bei „Shift4“ įkūrėjo sprendimą nesimokyti universitete ir vietoj to nuo pat paauglystės rinktis antreprenerio kelią.

    Kaip ir galima tikėtis, pažeriama ir daugiau išminties perlų:

    Tokio tipo integruoti mokėjimų sprendimai verslui daug labiau paplitę JAV nei Europoje. T. Lauberis tai aiškina tuo, jog Senajame žemyne istoriškai prioritetas buvo teikiamas vartotojo patirčiai ir saugumui, o ne pardavėjo poreikiams. Nors dabar taip nebėra ir integruoti mokėjimų sprendimai yra saugūs, Europa į šį traukinį vėluoja.
    ...
    Jis teigia, kad įprastiniai atsiskaitymo terminalai, kokius naudojame ir Lietuvoje, yra saugūs ir patrauklūs pirkėjams, bet galų gale „tai tėra banko terminalas“.
    „Jis nėra prijungtas prie kitų dalykų, kurių gali reikėti pardavėjui jo versle. Pavyzdžiui, lojalumo programų ar paskatų tam, kad sugrįžtum pas jį pirkti ateityje. Arba, pavyzdžiui, sistema, kuri naudojama jūsų mėsainio ar alaus užsakymui užfiksuoti bei jam nusiųsti į virtuvę, – ji veikia visiškai atskirai nuo terminalo, kuris jums paduodamas, sistemos“, – aiškina T. Lauberis.

    Iššifruokim ką direktorius vadina „pažanga“: pirmiausia, ne kliento patogumą ir saugumą, o verslo interesus. Būtent dėmesys kliento saugumui ir privatumui yra inovacija, o ne siautėjantys duomenų brokeriai ir nesustojantis kiekvieno žingsnio sekimas, neva pateisinantis agresyvų kišimąsi į žmonių privatų gyvenimą dėl kažin kokios menkutės lojalumo nuolaidos. Visiškas akių dūmimas :neocat_thumbsdown:

    JAV prireikė labai daug laiko išlipti iš magnetinių mokėjimo kortelių - būtent JAV dešimtmečiais atsilieka nuo Europos. Bet, žinoma, milijardieriai norėtų, kad visi manytų kitaip :neocat_glare:

    #verslas #mokėjimai #Lietuvoje #aktualijos #shift4

    #verslas #mokejimai #lietuvoje #aktualijos #shift4
  26. Gytis Repečka @[email protected] ·

    Labai nemėgstu samprotavimų, kad esą Europa technologiškai atsilieka nuo JAV. Melas ir paistalai! :neocat_foxmask:

    „Verslo žinios“ kalbina JAV kontorą „Shift4“, gaminančią mokėjimų sprendimus. Jos įkūrėjas milijardierius Jared Isaacman didžiuojasi užuot mokęsis universitete, pasukęs tiesiai į verslą, ir yra tiek apsėstas kosmoso reikalais, kad prie Trumpo net tapo NASA vadovu. Į panašią dūdą pučia dabartinis įmonės CEO Taylor Lauber:

    Jis su „Shift4“ įkūrėju „tėvų rūsyje“ dirbo dar paauglystėje ir, jo paties žodžiais, buvo „tuo idiotu, kuris nusprendė eiti mokytis į koledžą“. Taip T. Lauberis referuoja į jo ir J. Isaacmano išsiskyrusius kelius bei „Shift4“ įkūrėjo sprendimą nesimokyti universitete ir vietoj to nuo pat paauglystės rinktis antreprenerio kelią.

    Kaip ir galima tikėtis, pažeriama ir daugiau išminties perlų:

    Tokio tipo integruoti mokėjimų sprendimai verslui daug labiau paplitę JAV nei Europoje. T. Lauberis tai aiškina tuo, jog Senajame žemyne istoriškai prioritetas buvo teikiamas vartotojo patirčiai ir saugumui, o ne pardavėjo poreikiams. Nors dabar taip nebėra ir integruoti mokėjimų sprendimai yra saugūs, Europa į šį traukinį vėluoja.
    ...
    Jis teigia, kad įprastiniai atsiskaitymo terminalai, kokius naudojame ir Lietuvoje, yra saugūs ir patrauklūs pirkėjams, bet galų gale „tai tėra banko terminalas“.
    „Jis nėra prijungtas prie kitų dalykų, kurių gali reikėti pardavėjui jo versle. Pavyzdžiui, lojalumo programų ar paskatų tam, kad sugrįžtum pas jį pirkti ateityje. Arba, pavyzdžiui, sistema, kuri naudojama jūsų mėsainio ar alaus užsakymui užfiksuoti bei jam nusiųsti į virtuvę, – ji veikia visiškai atskirai nuo terminalo, kuris jums paduodamas, sistemos“, – aiškina T. Lauberis.

    Iššifruokim ką direktorius vadina „pažanga“: pirmiausia, ne kliento patogumą ir saugumą, o verslo interesus. Būtent dėmesys kliento saugumui ir privatumui yra inovacija, o ne siautėjantys duomenų brokeriai ir nesustojantis kiekvieno žingsnio sekimas, neva pateisinantis agresyvų kišimąsi į žmonių privatų gyvenimą dėl kažin kokios menkutės lojalumo nuolaidos. Visiškas akių dūmimas :neocat_thumbsdown:

    JAV prireikė labai daug laiko išlipti iš magnetinių mokėjimo kortelių - būtent JAV dešimtmečiais atsilieka nuo Europos. Bet, žinoma, milijardieriai norėtų, kad visi manytų kitaip :neocat_glare:

    #verslas #mokėjimai #Lietuvoje #aktualijos #shift4

    #shift4 #aktualijos #lietuvoje #mokejimai #verslas
  27. Gytis Repečka @[email protected] ·

    Labai nemėgstu samprotavimų, kad esą Europa technologiškai atsilieka nuo JAV. Melas ir paistalai! :neocat_foxmask:

    „Verslo žinios“ kalbina JAV kontorą „Shift4“, gaminančią mokėjimų sprendimus. Jos įkūrėjas milijardierius Jared Isaacman didžiuojasi užuot mokęsis universitete, pasukęs tiesiai į verslą, ir yra tiek apsėstas kosmoso reikalais, kad prie Trumpo net tapo NASA vadovu. Į panašią dūdą pučia dabartinis įmonės CEO Taylor Lauber:

    Jis su „Shift4“ įkūrėju „tėvų rūsyje“ dirbo dar paauglystėje ir, jo paties žodžiais, buvo „tuo idiotu, kuris nusprendė eiti mokytis į koledžą“. Taip T. Lauberis referuoja į jo ir J. Isaacmano išsiskyrusius kelius bei „Shift4“ įkūrėjo sprendimą nesimokyti universitete ir vietoj to nuo pat paauglystės rinktis antreprenerio kelią.

    Kaip ir galima tikėtis, pažeriama ir daugiau išminties perlų:

    Tokio tipo integruoti mokėjimų sprendimai verslui daug labiau paplitę JAV nei Europoje. T. Lauberis tai aiškina tuo, jog Senajame žemyne istoriškai prioritetas buvo teikiamas vartotojo patirčiai ir saugumui, o ne pardavėjo poreikiams. Nors dabar taip nebėra ir integruoti mokėjimų sprendimai yra saugūs, Europa į šį traukinį vėluoja.
    ...
    Jis teigia, kad įprastiniai atsiskaitymo terminalai, kokius naudojame ir Lietuvoje, yra saugūs ir patrauklūs pirkėjams, bet galų gale „tai tėra banko terminalas“.
    „Jis nėra prijungtas prie kitų dalykų, kurių gali reikėti pardavėjui jo versle. Pavyzdžiui, lojalumo programų ar paskatų tam, kad sugrįžtum pas jį pirkti ateityje. Arba, pavyzdžiui, sistema, kuri naudojama jūsų mėsainio ar alaus užsakymui užfiksuoti bei jam nusiųsti į virtuvę, – ji veikia visiškai atskirai nuo terminalo, kuris jums paduodamas, sistemos“, – aiškina T. Lauberis.

    Iššifruokim ką direktorius vadina „pažanga“: pirmiausia, ne kliento patogumą ir saugumą, o verslo interesus. Būtent dėmesys kliento saugumui ir privatumui yra inovacija, o ne siautėjantys duomenų brokeriai ir nesustojantis kiekvieno žingsnio sekimas, neva pateisinantis agresyvų kišimąsi į žmonių privatų gyvenimą dėl kažin kokios menkutės lojalumo nuolaidos. Visiškas akių dūmimas :neocat_thumbsdown:

    JAV prireikė labai daug laiko išlipti iš magnetinių mokėjimo kortelių - būtent JAV dešimtmečiais atsilieka nuo Europos. Bet, žinoma, milijardieriai norėtų, kad visi manytų kitaip :neocat_glare:

    #verslas #mokėjimai #Lietuvoje #aktualijos #shift4

    #verslas #mokejimai #lietuvoje #aktualijos #shift4
  28. EshaHaber @[email protected] ·

    News Haber EshaHaber Trabzonspor'a piyango! Kiralık ismin bonservisini alacaklar: Trabzonspor'un sezon başında Palermo'ya kiralık olarak gönderdiği Rayyan Baniya, İtalyan takımındaki performansıyla dikkatleri üzerine çekti.

    PALERMO BONSERVİSİNİ ALACAK

    Sabah Gazetesi'nde yer alan habere göre; Baniya'nın form grafiğinden memnun olan Serie B ekibi, İtalyan ve Türk vatandaşlığı da bulunan… eshahaber.com.tr/haber/trabzon EshaHaber.com.tr #Trabzonspor #RayyanBaniya #Palermo #transfer #futbol

    #futbol #transfer #palermo #rayyanbaniya #trabzonspor
  29. EshaHaber @[email protected] ·

    News Haber EshaHaber Trabzonspor'a piyango! Kiralık ismin bonservisini alacaklar: Trabzonspor'un sezon başında Palermo'ya kiralık olarak gönderdiği Rayyan Baniya, İtalyan takımındaki performansıyla dikkatleri üzerine çekti.

    PALERMO BONSERVİSİNİ ALACAK

    Sabah Gazetesi'nde yer alan habere göre; Baniya'nın form grafiğinden memnun olan Serie B ekibi, İtalyan ve Türk vatandaşlığı da bulunan… eshahaber.com.tr/haber/trabzon EshaHaber.com.tr #Trabzonspor #RayyanBaniya #Palermo #transfer #futbol

    #trabzonspor #rayyanbaniya #palermo #transfer #futbol
  30. EshaHaber @[email protected] ·

    MasterChef seyircisini şaşırtan ayrılık! Veda rüzgarları esiyor: Yarışmaya katıldığı ilk günden bu yana performansıyla isminden bahsettiren Furkan Kahraman'ın MasterChef serüvenine nokta koyuldu. Bilindiği üzere ana kadrodan sekizinci yarışmacı olarak MasterChef yolculuğuna adım atan Furkan Kahraman bu hafta rakiplerinin gerisinde kaldı. Jüri üyeleri tarafından yapılan ortak değerlendirmeler… eshahaber.com.tr/haber/masterc EshaHaber.com.tr #MasterChef #FurkanKahraman #yarışma #veda #eleme

    #masterchef #furkankahraman #yarisma #veda #eleme
Next page