AI repos for Speech and audio AI
Open-source AI projects built around Speech and audio AI or genuinely useful if you work with it, each with what it does for you. 23 repos, most relevant first.
-
whisper
Transcribe, translate and detect language in audio with six model sizes from CLI or Python.
-
VoiceStudio
Clone voices, dub video, transcribe and make audiobooks locally instead of using a cloud service.
-
voicebox
Clone voices, generate speech with seven TTS engines and dictate with Whisper, all locally.
-
VoxCPM
Generate 48kHz speech in 30 languages, design voices from text descriptions and clone voices from short clips.
-
VoiceInk
Native macOS dictation using local speech-to-text models, with readable source you can build yourself.
-
meetily
Transcribe meetings live and locally with Whisper or Parakeet, mixing mic and system audio.
-
supertonic
Offline 31-language text-to-speech with a 99M model that outputs 44.1kHz WAV without a GPU.
-
FluidVoice
Run several on-device speech-to-text models for system-wide dictation on macOS, with optional AI cleanup.
-
openai-realtime-meeting-assistant
Worked example of voice input through OpenAI's Realtime API over WebRTC driving app actions.
-
omi
Real-time transcription of conversations with summaries and action items.
-
earshot
Describe and match sounds, and synthesize audio in code, locally on CPU.
-
doodle-note
Transcribes microphone and call audio on-device, with no bot joining the meeting.
-
tldr-radio
Uses the local Kokoro text-to-speech model to read newsletters aloud, with chapters and selectable voices.
-
transformers
Run speech and audio models through the same pipeline API.
-
MoneyPrinterTurbo
Generates voiceover and subtitles automatically for short videos using text-to-speech.
-
openai-agents-python
Add realtime and voice agents to your agent workflows.
-
gallery
Audio Scribe transcribes speech entirely on the device.
-
screenpipe
Continuously transcribes your computer's audio locally, making meetings and calls searchable.
-
claude-video
Pulls captions or falls back to transcription so the agent gets a timestamped transcript.
-
needle
Bundles Whistle speech-to-text for on-device voice input.
-
ACE-Step-1.5
Generates full songs with lyrics and handles covers, repainting and audio stem separation.
-
clicky
Combines push-to-talk voice input with spoken replies in a screen-aware assistant.
-
openshorts
Generates captions and dubbing from long-video audio.