#speechtotext — Public Fediverse posts

Vocalinux is an open-source voice dictation app for Linux that works entirely offline.

It supports Whisper.cpp, OpenAI Whisper, and VOSK, letting you dictate into almost any application without sending your audio to the cloud. It also includes GPU acceleration, customizable hotkeys, and Wayland/X11 support.

More details: https://digitalescapetools.com/tools/tool.html?id=vocalinux

#OpenSource #Linux #Privacy #SpeechToText #Whisper #Productivity

Vocalinux is an open-source voice dictation app for Linux that works entirely offline.

It supports Whisper.cpp, OpenAI Whisper, and VOSK, letting you dictate into almost any application without sending your audio to the cloud. It also includes GPU acceleration, customizable hotkeys, and Wayland/X11 support.

More details: https://digitalescapetools.com/tools/tool.html?id=vocalinux

#OpenSource #Linux #Privacy #SpeechToText #Whisper #Productivity

Vocalinux is an open-source voice dictation app for Linux that works entirely offline.

It supports Whisper.cpp, OpenAI Whisper, and VOSK, letting you dictate into almost any application without sending your audio to the cloud. It also includes GPU acceleration, customizable hotkeys, and Wayland/X11 support.

More details: https://digitalescapetools.com/tools/tool.html?id=vocalinux

#OpenSource #Linux #Privacy #SpeechToText #Whisper #Productivity

Vocalinux is an open-source voice dictation app for Linux that works entirely offline.

It supports Whisper.cpp, OpenAI Whisper, and VOSK, letting you dictate into almost any application without sending your audio to the cloud. It also includes GPU acceleration, customizable hotkeys, and Wayland/X11 support.

More details: https://digitalescapetools.com/tools/tool.html?id=vocalinux

#OpenSource #Linux #Privacy #SpeechToText #Whisper #Productivity

Vocalinux is an open-source voice dictation app for Linux that works entirely offline.

It supports Whisper.cpp, OpenAI Whisper, and VOSK, letting you dictate into almost any application without sending your audio to the cloud. It also includes GPU acceleration, customizable hotkeys, and Wayland/X11 support.

More details: https://digitalescapetools.com/tools/tool.html?id=vocalinux

#OpenSource #Linux #Privacy #SpeechToText #Whisper #Productivity

@itsfoss How does it compare to #SpeedofSound app by @zugaldia?

https://flathub.org/en/apps/io.speedofsound.SpeedOfSound

#voicetyping #speechtotext #dictationsoftware #accessibility #a11ty #linuxaccessibility #accessibilityonlinux

@itsfoss How does it compare to #SpeedofSound app by @zugaldia?

https://flathub.org/en/apps/io.speedofsound.SpeedOfSound

#voicetyping #speechtotext #dictationsoftware #accessibility #a11ty #linuxaccessibility #accessibilityonlinux

Аналог Plaud на iPhone: эксперименты с локальной транскрибацией ASR на iOS

Мы делаем приложение Memo: есть куча гаджетов и сервисов вроде Plaud — это когда ты платишь за отдельную «умную» коробочку-диктофон, которая записывает встречу и сама делает из неё протокол. Зачем покупать отдельную железку, если в кармане уже есть iPhone? Давайте соберём такой же «умный протоколщик» прямо на телефоне.

https://habr.com/ru/articles/1054566/

#iOS #распознавание_речи #ASR #speechtotext #ondevice_ML #локальные_модели #транскрибация #SpeechAnalyzer #Apple_Neural_Engine #разработка_под_iOS

Hi It's me asking for accessibility help again! Anyone know of a good simple speech to text program for *Windows*? It needs to handle long ramblings so I can just talk out my drafts by my desktop - (no sub fees,no sneaky llm training etc)

Don't suggest Linux because yes I will be getting something for Linux for my laptop very soon - It's already being sorted!

Boosts appreciated!

#accessibility #speechtoText

Voicebox is a free and open-source AI voice studio that runs entirely on your computer.

It lets you clone voices, generate speech, transcribe audio with Whisper, and use voice dictation without sending recordings to the cloud. It also supports multiple TTS engines and MCP integration for AI assistants.

More details: https://digitalescapetools.com/tools/tool.html?id=voicebox

#OpenSource #AI #Privacy #SelfHosted #Linux #Windows #macOS #SpeechToText #TextToSpeech

Voicebox is a free and open-source AI voice studio that runs entirely on your computer.

It lets you clone voices, generate speech, transcribe audio with Whisper, and use voice dictation without sending recordings to the cloud. It also supports multiple TTS engines and MCP integration for AI assistants.

More details: https://digitalescapetools.com/tools/tool.html?id=voicebox

#OpenSource #AI #Privacy #SelfHosted #Linux #Windows #macOS #SpeechToText #TextToSpeech

Voicebox is a free and open-source AI voice studio that runs entirely on your computer.

It lets you clone voices, generate speech, transcribe audio with Whisper, and use voice dictation without sending recordings to the cloud. It also supports multiple TTS engines and MCP integration for AI assistants.

More details: https://digitalescapetools.com/tools/tool.html?id=voicebox

#OpenSource #AI #Privacy #SelfHosted #Linux #Windows #macOS #SpeechToText #TextToSpeech

Voicebox is a free and open-source AI voice studio that runs entirely on your computer.

It lets you clone voices, generate speech, transcribe audio with Whisper, and use voice dictation without sending recordings to the cloud. It also supports multiple TTS engines and MCP integration for AI assistants.

More details: https://digitalescapetools.com/tools/tool.html?id=voicebox

#OpenSource #AI #Privacy #SelfHosted #Linux #Windows #macOS #SpeechToText #TextToSpeech

Accessibility help!
Does anyone know of a good simple free *Speech to Text* app for android phones (only needs to handle English) that isn't gonna be littered with ads or sneaky LLM training or anything like that?

EDIT. Thank you very much. I have found one app that is amazing. It's called Futo and I will be using that from now.

No more help needed. Exclamation mark. 😂

#Accessibility #SpeechToText #speech_to_text #Help

If you spend hours writing emails, documents, code, or notes, typing everything can quickly become exhausting. FluidVoice makes dictation feel much more natural by turning your voice into text with fast, accurate speech recognition and optional on-device AI enhancements.

Its built with a local-first approach. Most speech models run directly on your Mac

#macos #AI #opensource #speechtotext #software #developers

https://firethering.com/fluidvoice-ai-voice-to-text-dictation-app-for-macos/

If you spend hours writing emails, documents, code, or notes, typing everything can quickly become exhausting. FluidVoice makes dictation feel much more natural by turning your voice into text with fast, accurate speech recognition and optional on-device AI enhancements.

Its built with a local-first approach. Most speech models run directly on your Mac

#macos #AI #opensource #speechtotext #software #developers

https://firethering.com/fluidvoice-ai-voice-to-text-dictation-app-for-macos/

On the nights I can't get to sleep it is usually because a thought gets in my head and won't shut up.
Tonight I turned my browser back on to check the expected future cataract operation.

Having been told within 18 months, I was thinking about plans to cope for next year. But having moved to noai duckduckgo I just found some real research based on several studies.
It could be 3 months, and about 50% get cataracts that need surgery with 6 months.
I need to go from thinking about planning to doing it, especially finding apps that actually work for dictation.

I've tried a few in the past, but too many are built for American male voices. The last one refused to type 'that' for me even though I have a standard English accent.

My first experience with voice activated cameras for video meeting was a good example. The camera moved to focus on the last man who spoke and completely ignored every woman in the room. And while Dragon worked OK, they no longer do personal licences.

#SpeechToText

On the nights I can't get to sleep it is usually because a thought gets in my head and won't shut up.
Tonight I turned my browser back on to check the expected future cataract operation.

Having been told within 18 months, I was thinking about plans to cope for next year. But having moved to noai duckduckgo I just found some real research based on several studies.
It could be 3 months, and about 50% get cataracts that need surgery with 6 months.
I need to go from thinking about planning to doing it, especially finding apps that actually work for dictation.

I've tried a few in the past, but too many are built for American male voices. The last one refused to type 'that' for me even though I have a standard English accent.

My first experience with voice activated cameras for video meeting was a good example. The camera moved to focus on the last man who spoke and completely ignored every woman in the room. And while Dragon worked OK, they no longer do personal licences.

#SpeechToText

@d1 @housepanther Handy/Parler ?
https://github.com/Melvynx/Parler
#speechToText

@d1 @housepanther Handy/Parler ?
https://github.com/Melvynx/Parler
#speechToText

I've just added a very early preview of “meetings” to Dictator on the Mac (dictator.robgough.net). Right now, the recommendation is 32GB RAM, but with that you can get fully offline meeting summaries. It’s still going to need tuning, but it works quite well! #buildinpublic #macos #speechtotext

Как мы превращаем звонок риэлтора в карточку лида за 15 секунд: ИИ-автолид изнутри

Риэлтор за рулём. Звонит собственник трёшки на Соколе: “Видел ваше объявление, хочу обсудить продажу”. Двадцать минут живого разговора - район, перепланировка, срочность, вилка по цене. Разговор кончается, риэлтор едет на показ, к вечеру у него ещё пять звонков. Утром он помнит, что “был кто-то по трёшке”, но не помнит ни имени, ни цены, ни телефона. Лид потерян не потому, что плохо отработали, а потому, что между звонком и CRM стоит человек с памятью и руками, которые в этот момент держат руль.

https://habr.com/ru/articles/1044520/

#распознавание_речи #deepgram #llm #speechtotext #crm #автоматизация #пайплайн #недвижимость

Ich: "Welches E-Mailprogramm hast Du denn?"

Schwager: "Fandabööd."

#Original
#IT-Profession
#SpeechToText

Ich: "Welches E-Mailprogramm hast Du denn?"

Schwager: "Fandabööd."

#Original
#IT-Profession
#SpeechToText

Dictator is live on the iOS App Store!

Dictator is now available for those who want to try out talking instead of typing without needing to pay. Essentially the likes of wisprflow but for free!

https://apps.apple.com/gb/app/dictator-voice-typing/id6772823676

#speechtotext #dictation #iphone

Как я решил проблему русской диктовки для ИИ

По мере погружения в ИИ и вайб‑кодинг, я столкнулся с одним неудобным моментом — отсутствием возможности диктовать на русском языке в некоторых программах. И если OpenAI в своем приложении позаботились об этом, то в Anthropic такой возможности на тот момент просто не оказалось. А мне уже так понравилось, откинувшись на спинку кресла с чашкой чая, надиктовывать промпты без клавиатуры. Но я быстро нашел выход, хоть и костыльный — просто диктовать свой текст в окошке GPT, потом копировать его и вставлять в Claude. Вроде несложно, но и удобным этот метод я бы не назвал. И я задумался, как этот процесс оптимизировать. И какая же идея могла прийти в голову в 3 часа ночи человеку, который полжизни занимается программированием? Ну конечно же — разработать свое приложение. Посоветовавшись с Claude и GPT, я набросал небольшой план и приступил к разработке. Поскольку я работаю на macOS, то для начала не стал заморачиваться с мультиплатформенностью и решил делать все на Swift.

https://habr.com/ru/articles/1039248/

#AI #OpenAI #Claude #Whisper #speechtotext #диктовка #voice_input #Apple_Silicon

Schwerpunkt 1
Lokales Speech to Text in Linux Mint einrichten: Auf Knopfdruck (beliebiger Shortkey) das Diktat starten & beenden, in jeder beliebigen Anwendung. Keine Cloud, kein mithörender Datenkrake, nur 4-6 GB im RAM.

Schwerpunkt 2
Wie mir KI (GPT) geholfen hat, das alles hinzubekommen, inkl. kompletter Projektdokumentation.

https://blog.derbrumme.de/lokales-speech-to-text-in-linux-mint-einrichten/

#Linux #OpenSource #SpeechToText #STT #Vosk #Privacy #SelfHosting #KI #AI #NerdDictation

Schwerpunkt 1
Lokales Speech to Text in Linux Mint einrichten: Auf Knopfdruck (beliebiger Shortkey) das Diktat starten & beenden, in jeder beliebigen Anwendung. Keine Cloud, kein mithörender Datenkrake, nur 4-6 GB im RAM.

Schwerpunkt 2
Wie mir KI (GPT) geholfen hat, das alles hinzubekommen, inkl. kompletter Projektdokumentation.

https://blog.derbrumme.de/lokales-speech-to-text-in-linux-mint-einrichten/

#Linux #OpenSource #SpeechToText #STT #Vosk #Privacy #SelfHosting #KI #AI #NerdDictation

Speech-to-LaTeX: распознавание математических выражений и предложений в LaTeX

Представьте семинар у физиков или математиков. Идёт автоматическая запись лекции, а затем распознавание речи в аккуратный текст. В большинстве мест современные ASR-системы справятся неплохо. Но значительная часть такой записи будет состоять из фраз вроде «интеграл от икс в квадрате до бесконечности», «сумма по i от единицы до n» или «производная по t от функции f». Формально голос может быть распознан правильно. В расшифровке даже могут появляться отдельные символы вроде + , π или x . Но если человек произносит длинную формулу, результат почти всегда превращается в линейную фразу, читать которую физически больно. Хочется другого: чтобы система сразу понимала, где обычный текст, где математическое выражение, и выдавала не «один делить на икс плюс два», а корректный LaTeX-код, например, \frac{1}{x+2} или \frac{1}{x}+2 , в зависимости от смысла. Эта задача называется Speech-to-LaTeX или S2L: преобразование озвученных математических выражений и предложений в формальную LaTeX-запись. В отличие от обычного speech-to-text, здесь нужно распознать не только слова, но и структуру: дроби, индексы, степени, пределы, суммы, интегралы, скобки, вложенные выражения и границы формул. Например, фраза «два делить на пи» в обычной расшифровке может остаться как «2 делить на π». Но в LaTeX она должна стать \frac{2}{\pi} . Именно такой формат нужен для статей, учебников, конспектов, Overleaf и других LaTeX-редакторов. Несмотря на прогресс в automatic speech recognition (ASR), задача прямого преобразования озвученной математики в LaTeX долго оставалась почти неразработанной. Более того, нормальных открытых датасетов с человеческими аудиозаписями для такой задачи практически не было. В нашей работе мы попытались закрыть этот пробел: собрали открытый двуязычный датасет и сравнили несколько подходов к Speech-to-LaTeX. В статье , которую мы представили на ICLR 2026, описан датасет из более чем 66 тысяч человеческих аудиозаписей и 571 тысячи синтетических аудиозаписей на английском и русском языках.

https://habr.com/ru/companies/airi/articles/1036562/

#ASR #llmмодели #latex #speechtotext #speechtolatex

murmure - Fully local, private and cross platform Speech-to-Text with #LLM Post-processing:

Website:
https://www.murmure.app/

Repo:
https://github.com/Kieirra/murmure

This looks awesome! :awesome:

- every operation happens locally on device
- supports 25 European languages
- uses NVIDIA's Parakeet model for transcription
- build with #Tauri and #Rust

#SpeechToText #LLMs #Privacy #RustLang

Ok, est ce que je viens d'être bluffé par la saisie vocal de #Outspoke ?

Modèle hors ligne, application open-source, support du français et autres langues européennes, intégration avec le clavier, nettoyage des "hum, heu..."

Je continu ?
C'est la bonne découverte ! https://apt.izzysoft.de/fdroid/index/apk/dev.brgr.outspoke
#stt #opensource #keyboard #speechtotext

Ok, est ce que je viens d'être bluffé par la saisie vocal de #Outspoke ?

Modèle hors ligne, application open-source, support du français et autres langues européennes, intégration avec le clavier, nettoyage des "hum, heu..."

Je continu ?
C'est la bonne découverte ! https://apt.izzysoft.de/fdroid/index/apk/dev.brgr.outspoke
#stt #opensource #keyboard #speechtotext

Ok, est ce que je viens d'être bluffé par la saisie vocal de #Outspoke ?

Modèle hors ligne, application open-source, support du français et autres langues européennes, intégration avec le clavier, nettoyage des "hum, heu..."

Je continu ?
C'est la bonne découverte ! https://apt.izzysoft.de/fdroid/index/apk/dev.brgr.outspoke
#stt #opensource #keyboard #speechtotext

Ok, est ce que je viens d'être bluffé par la saisie vocal de #Outspoke ?

Modèle hors ligne, application open-source, support du français et autres langues européennes, intégration avec le clavier, nettoyage des "hum, heu..."

Je continu ?
C'est la bonne découverte ! https://apt.izzysoft.de/fdroid/index/apk/dev.brgr.outspoke
#stt #opensource #keyboard #speechtotext

Ok, est ce que je viens d'être bluffé par la saisie vocal de #Outspoke ?

Modèle hors ligne, application open-source, support du français et autres langues européennes, intégration avec le clavier, nettoyage des "hum, heu..."

Je continu ?
C'est la bonne découverte ! https://apt.izzysoft.de/fdroid/index/apk/dev.brgr.outspoke
#stt #opensource #keyboard #speechtotext

@yv_pic @gaunardemmanuel @brigit_komit Chez-moi c'est bloqué par les dispositifs de protection de la vie privée…
Personnellement si cela pouvait tourner sans navigateur j'aimerais beaucoup.
#speechToText #dictée #reconnaissanceDeLaParole

Почему Cluely и другие плохо слышат русских айтишников: разбор того, как Whisper ломается и что мы сделали с этим

В январе я купил подписки на Cluely, Final Round AI и Sensei. Хотел посмотреть как они справляются с русским айти-собесами, раз уж все три заявляют о поддержке русского. Подключил по очереди к тестовому звонку в Телемосте (сомневаюсь, что платформа имела роль, но все же), прогнал одну и ту же запись: Senior Python backend разработчик, 45 минут, стек FastAPI + PostgreSQL + Kafka + Kubernetes. Обычный русский спикер, если важно - из Москвы, с речью проблем не было, нормальный микрофон Все три выдали транскрипт и все три провалились, как неожиданно.. "Кафка" в половине случаев становилась "как-то" или "кофта". "Кубернетис" превращался в "губер нет тест". "Сабскрайбер патерн" - в "саб скрайп патерн". "Middleware для CSRF" - "мидл-вер для си эс эр эф" - это еще норм Проблема не в том, что человек говорил по-русски, и не в том, что Whisper не умеет русский (сноска: хорошо не умеет). Whisper умеет русский нормально, около 9.8% WER на Common Voice. Проблема в другом: русскоязычный айтишник не говорит ни на чистом русском, ни на чистом английском. Он говорит на гибриде: русская грамматика плюс английские термины плюс своеобразное произношение этих терминов плюс местами свой жаргон вроде "гошечки" и "крудошлёпа" Этот гибрид ни один из популярных STT не держит. Потому что его в тренировочных данных почти нет Разбираю ниже, как устроена эта проблема, что с ней делают конкуренты (почти ничего), и что сделали мы

https://habr.com/ru/articles/1026778/

#Whisper #STT #speechtotext #finetuning #LoRA #ASR #NLP #распознавание_речи #русский_язык #codeswitching

Scrivere senza tastiera su Linux è possibile 🎙️ Scopri Speed of Sound e il riconoscimento vocale offline #Linux #OpenSource #Whisper #SpeechToText #Software

https://www.linuxeasy.org/speed-of-sound-porta-la-dettatura-vocale-su-linux/?utm_source=mastodon&utm_medium=jetpack_social

Fedi folks! I'm looking for a speech to text app for my android phone or alternatively software for my Windows computer that turns voice memos into text. It needs to work offline, ideally #opensource and I'm willing to pay for it.

How would you go about this?

I'll be out in the field surveying biotopes soon and would like to record voice memos instead of scribbling illegible notes in the rain.

#fedirecommendations #voicememo #speechtotext #recommendations #fieldbotany #fieldbiology

Fedi folks! I'm looking for a speech to text app for my android phone or alternatively software for my Windows computer that turns voice memos into text. It needs to work offline, ideally #opensource and I'm willing to pay for it.

How would you go about this?

I'll be out in the field surveying biotopes soon and would like to record voice memos instead of scribbling illegible notes in the rain.

#fedirecommendations #voicememo #speechtotext #recommendations #fieldbotany #fieldbiology

https://www.omgubuntu.co.uk/2026/04/speed-of-sound-linux-voice-typing-app

"Speed of Sound" is an Open Source Linux app that allows speech-to-text in any focused window, using the cutting edge Whisper speech recognition model

#OpenSource #Linux #SpeechToText #STT #SpeedOfSound

Почему одного Whisper оказалось недостаточно и как мы создали полноценный сервис распознавания речи

Всем привет! Меня зовут Наталья, я инженер машинного обучения в ЮMoney. Мы уже писали о том, как транскрибируем аудио с внутренних созвонов в текст. Прошёл год, и задача выросла: помимо созвонов мы решили транскрибировать все звонки службы поддержки, а также создать удобный интерфейс для работы с аудио и текстом. В этой статье расскажу, как нам удалось реализовать всё это, и при этом повысить качество распознавания и сохранить процесс внутри корпоративного контура. Мы протестировали различные решения и теперь делимся опытом, чтобы другие команды могли быстрее внедрять проверенные подходы и избегать распространённых ошибок.

https://habr.com/ru/companies/yoomoney/articles/1012870/

#распознавание_речи #speechtotext #whisper #аудиообработка #диаризация #речевая_аналитика #машинное_обучение #vad

@jenniferplusplus

You stated: <<What it actually says is "AI is _permitted_ to make mistakes; _you are liable_ for the results, whether you check them or not". Except "you" is generally not even the person building, installing, or even using the AI. It's the person the AI is used on.>>

Way back in the early 2000s, there was a system called "Dragon Dictate". The goal was to eliminate #human #transcriptionists with automated speech-to-text (sound familiar?) The system had to be trained on your voice and vocabulary. Once properly trained it could do a pretty good job, I'll guess 95-98%. It was better suited to output that was stereotyped (mostly the same), and structured (such as radiology reports and operative notes).

Regardless of how the note/report was generated, the professional who spoke the words had a obligation to at least scan the output and sign it (yes, with an ink pen!). Once signed it became part of the "legal medical record" open to misinterpretation, copying, lawsuits, etc. etc.

Once Dragon Dictate became routine (and they fired all the transcriptionists) I started to notice this little #disclaimer at the bottom:

"If portions of this note are confusing or indecipherable please feel free to call me with questions or concerns." Sounds a lot like #AI to me! I polite way to summarize this is:

👉 They were trying to force me to be their copy-editor. 👈

It cast the entire content in doubt.

Consider for a moment the difference between saying "The scan does not show cancer." and "The scan does show cancer." That "not" is doing a lot of work, and is very easy to miss when you're talking fast and never intend to read your own note ever again.

More subtle is the grammatical error in the first sentence. "This note was #dictated using Dragon text to speech recognition software." Either they changed their product name to "Dragon Text", in which case the capitalization is off. Or they transposed words and it should read "speech to text" or "speech recognition" with no text.

👉 In other words, they didn't even proof-read their own disclaimer! 😱

#MedicalRecords #Medicine #SpeechToText #Liability #Risk #SignalToNoise

@jenniferplusplus

You stated: <<What it actually says is "AI is _permitted_ to make mistakes; _you are liable_ for the results, whether you check them or not". Except "you" is generally not even the person building, installing, or even using the AI. It's the person the AI is used on.>>

Way back in the early 2000s, there was a system called "Dragon Dictate". The goal was to eliminate #human #transcriptionists with automated speech-to-text (sound familiar?) The system had to be trained on your voice and vocabulary. Once properly trained it could do a pretty good job, I'll guess 95-98%. It was better suited to output that was stereotyped (mostly the same), and structured (such as radiology reports and operative notes).

Regardless of how the note/report was generated, the professional who spoke the words had a obligation to at least scan the output and sign it (yes, with an ink pen!). Once signed it became part of the "legal medical record" open to misinterpretation, copying, lawsuits, etc. etc.

Once Dragon Dictate became routine (and they fired all the transcriptionists) I started to notice this little #disclaimer at the bottom:

"If portions of this note are confusing or indecipherable please feel free to call me with questions or concerns." Sounds a lot like #AI to me! I polite way to summarize this is:

👉 They were trying to force me to be their copy-editor. 👈

It cast the entire content in doubt.

Consider for a moment the difference between saying "The scan does not show cancer." and "The scan does show cancer." That "not" is doing a lot of work, and is very easy to miss when you're talking fast and never intend to read your own note ever again.

More subtle is the grammatical error in the first sentence. "This note was #dictated using Dragon text to speech recognition software." Either they changed their product name to "Dragon Text", in which case the capitalization is off. Or they transposed words and it should read "speech to text" or "speech recognition" with no text.

👉 In other words, they didn't even proof-read their own disclaimer! 😱

#MedicalRecords #Medicine #SpeechToText #Liability #Risk #SignalToNoise

Efficient Real-Time Speech-to-Text Transcription #Speechtotext #Realtimetranscription #Voicecommands #Textinput #Oculus #Unity #AssetStore

https://u3dn.com/packages/efficient-real-time-speech-to-text-transcription-265013

Как я снизил WER с 33% до 3.3% для русской речи на CPU: сравнение GigaAM, Whisper и Vosk

За два месяца я перепробовал три ASR-движка, шесть моделей Whisper, адаптивное чанкование, T5-коррекцию и ансамблевое голосование — и большая часть идей оказалась тупиком. В статье — подробный разбор шести тупиков и одной находки: почему GigaAM от Сбера на обычном CPU показывает 3.3% WER на русском, обходя Whisper large-v3-turbo на RTX 4090 (7.9%) в 2.4 раза. С бенчмарками, кодом и честными оговорками.

https://habr.com/ru/articles/1002260/

#speechtotext #gigaam #whisper #vosk #onnx #распознавание_речи #WER #голосовой_ввод #ASR #python

I want to hear from someone who uses speech to text to write - how do you do it? Do you do something while you are speaking? What app do you use, how do you use it, and where does the text go? Where do you store what you dictated?

I am very curious because I once upon a time though I'd try it but never got to it. (^_^)* Would be worthy to give it a try while I am painting and see if my brain can do that at the same time.

#writing #speechtotext #dictating #writer #author

I want to hear from someone who uses speech to text to write - how do you do it? Do you do something while you are speaking? What app do you use, how do you use it, and where does the text go? Where do you store what you dictated?

I am very curious because I once upon a time though I'd try it but never got to it. (^_^)* Would be worthy to give it a try while I am painting and see if my brain can do that at the same time.

#writing #speechtotext #dictating #writer #author

I want to hear from someone who uses speech to text to write - how do you do it? Do you do something while you are speaking? What app do you use, how do you use it, and where does the text go? Where do you store what you dictated?

I am very curious because I once upon a time though I'd try it but never got to it. (^_^)* Would be worthy to give it a try while I am painting and see if my brain can do that at the same time.

#writing #speechtotext #dictating #writer #author