home.social

#voicesynthesis — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #voicesynthesis, aggregated by home.social.

fetched live
  1. Some improvements to the concatenation, prosody is still missing.

    Here is a well known phrase by SCP 079.

    The audio contains the same phrase first performed by Dr. Sbaitso TTS and the by Godot reimplementation.

    #TTS #DrSbaitso #VoiceSynthesis #TextToSpeech #079 #SCP079 #SCP #Godot

  2. Dr. Sbaitso compared to my reimplementation in Godot (Sbaitso first) :computer_explorer: :pc_color:

    Implemented: basic waveform concatenation
    Missing: Interpolation, pitch control, prosody, text to phonemes

    Im very happy with the progress, will be great to be able to run the voice without needing emulation.

    #TTS #DrSbaitso #VoiceSynthesis #TextToSpeech #079 #SCP079

  3. What I've learned so far while reverse engineering Dr Sbaitso's voice:
    - Reverse engineering is hard

    Also, the voice was made by very clever people. It's optimized to sound as good as possible, while consuming very few resources.

    Progress after 5 days: 10%

    #TTS #DrSbaitso #VoiceSynthesis #TextToSpeech

  4. Germany gets a new AI call assistant from Deutsche Telekom that works straight from the cellular network—no app required. Powered by ElevenLabs’ voice synthesis, it can translate languages on the fly. Unveiled at Mobile World Congress, it shows how open‑source‑friendly AI can reshape everyday calls. Curious how it works? #MagentaAI #DeutscheTelekom #ElevenLabs #VoiceSynthesis

    🔗 aidailypost.com/news/magenta-a

  5. Designing with the KT142C voice chip? Remember, its user-accessible memory is 320KB. If your audio files are larger, you'll need a different solution.

    The KT142F chip allows for an external Flash memory, letting you scale your voice storage capacity as needed. The trade-off is a slight increase in component cost and board space, but you gain total flexibility.

    linkedin.com/pulse/built-in-32

    #OpenHardware #Embedded #Electronics #VoiceSynthesis #PCBDesign #KT142 #Maker

  6. Ever wondered how AI voices are becoming so human-like? We've moved beyond simple "voice packs" (pre-recorded clips) to true AI "voice clones" that generate new speech from text.

    The magic is in the details: AI models learn a voice's unique pitch and cadence. The secret sauce? "Emotional tuning," which adds happiness, sadness, or empathy to the performance. It's a game-changer for accessibility and content creation. #AIVoice #VoiceSynthesis #Tech

  7. Project update: LingoFreq #Mandarin

    Generates themed example sentences from language word-frequency list (Chinese HSK). Builds a printable book & (soon) voice audio files. This theme is "technology".

    We'll be building this for #English & #Spanish.

    Tech: #Python, OpenAI ChatGPT, ElevenLabs #VoiceSynthesis (coming soon), SQLite database, Jinja templates

    Source code:
    - Sentence builder: codeberg.org/jro/LingoFreq-app
    - Book builder: codeberg.org/jro/LingoFreq-app

    #China

  8. Project update: LingoFreq - language learning tool

    Goal: Create voice audio files of high-frequency Mandarin words & example sentences

    Loads high-frequency #Mandarin #Chinese words from HSK CSV file into an #SQLite #database, using #Python. Then we loop over all the Mandarin words, and call to OpenAI ChatGPT to create simple example sentences. Next, we'll use ElevenLabs #VoiceSynthesis text-to-speech to vocalize the example sentences as audio files.

    Code: codeberg.org/jro/LingoFreq-app

    #China

  9. #Google, your automated service already told me that you're an automated service confirming our business hours. Don't patronize me by pretending to be a human. Why the bleep are you inserting artificial "um" and "sorry" and the like into your sentences? It doesn't fool me or make me feel better. #ai #voiceSynthesis

  10. @mark I've been doing digging into #openSource text to speech lately. The best overall I have found is Piper (github.com/OHF-Voice/piper1-gpl), but being honest - the voices are not great.

    To those in the know - Is there a ranked repository of better voices somewhere? How easy is it to create your own voices? - Using this for building language learning tools (#English, #Spanish, #Mandarin). Other TTS tools to consider?

    #VoiceSynthesis

  11. #OpenInvite #OpenSource #CodeJam @ #Medein
    Universidad #EAFIT / Bloque 20 pavilion / Behind Laboratorio de #Cafe. Next to Bloque 19 / #MakerSpace. We are learning & building with #Python, #FastAPi, #VoiceSynthesis.

    #NewbieFriendly

    Come by. Bring long power extension cables if you got'm!

  12. OK, this is probably a rather long shot, but does anyone know of a voice synthesis model that is open-weights or ideally even open-source, and capable of producing intonation (specifically, something like rap lyrics)? #text2speech #voicesynthesis #ai

  13. Google beats OpenAI to wide release of interruptible AI voice chat mode - Enlarge / The Google Gemini logo. (credit: Google)

    On Thursday... - arstechnica.com/?p=2049707 #advancedvoicemode #machinelearning #voicesynthesis #googlegemini #aivoicechat #geminilive #chatgpt #chatgtp #biz#google #openai #ai

  14. New Episode: hpr4188 :: Re: HPR4172 Comment by Ken Fallon

    Hosted by Archer72 on 2024-08-21 is flagged as Clean and is released under a CC-BY-SA license.

    Tags: #tts, #TextToSpeech, #VoiceSynthesis, #accessibility.

    hackerpublicradio.org/eps/hpr4

  15. ChatGPT Advanced Voice Mode impresses testers with sound effects, catching its breath - Enlarge / A stock photo of a robot whispering to a man. (credit: Andrey... - arstechnica.com/?p=2040213 #largelanguagemodels #advancedvoicemode #aivoicegenerators #machinelearning #audiosynthesis #voicesynthesis #textsynthesis #chatgpt #chatgtp #biz#openai #ai

  16. New Episode: hpr4172 :: Re: hpr4072 Piper voice synthesis

    Hosted by Archer72 on 2024-07-30 is flagged as Clean and is released under a CC-BY-SA license.

    Tags: #tts, #TextToSpeech, #VoiceSynthesis, #accessibility.

    hackerpublicradio.org/eps/hpr4