speech-recognition

huggingface/transformers

A unified model-definition framework spanning text, vision and audio — Transformers acts as the shared definition layer for models like DeepSeek, Gemma, GLM and Qwen, supporting both inference and training across text, vision, audio and multimodal tasks. It suits researchers and engineers who want one consistent API and access to pretrained checkpoints from the Hub instead of per-architecture code.

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

m-bain/whisperX

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)

abus-aikorea/voice-pro

One Gradio UI for TTS, Voice Cloning, Whisper, and Dubbing — Voice-Pro bundles Edge-TTS/Kokoro TTS, zero-shot voice cloning (E2/F5-TTS, CosyVoice), Whisper-based transcription, YouTube download, Demucs vocal isolation, and multilingual translation into one Gradio web app. It targets creators and developers who want an end-to-end pipeline for subtitling, dubbing, and podcast/audiobook production without stitching separate tools together.

Gradio WebUI for creators and developers, featuring key TTS (Edge-TTS, kokoro) and zero-shot Voice Cloning (E2 & F5-TTS, CosyVoice), with Whisper audio processing, YouTube download, Demucs vocal isolation, and multilingual translation.

QwenAudio/qwen-audio-agent

Keeping coding Agents talking while they work, not just answering — A realtime voice runtime that lets AI Agents stay present during long tasks—speaking while they think, call tools, or code, then announcing when work is done. v2.0 adds unified ACP/A2A backend integration, broader voice/video model support, memory, and desktop/mobile conversation panels, aimed at developers building voice-driven coding agents.

A realtime voice runtime that keeps Agents talking, working, and present. Real-time Voice Runtime for AI Agents

sgl-project/sglang-omni

A SGLang spin-off dedicated to serving speech and omni models — Built on the SGLang project, SGLang-Omni focuses on serving TTS, ASR and unified multimodal models with OpenAI-compatible APIs, streaming and distributed inference on CUDA. Useful for teams who need production-grade speech and multimodal serving without building custom inference stacks.

High Performance Serving Runtime For Interactive Super Intelligence System

yuxino/Mimi

Cross-platform live subtitles that translate system or mic audio in real time — Mimi is a Tauri/Rust desktop and Android app that captures computer or microphone audio and overlays real-time translated subtitles. It supports macOS, Windows, Linux (PipeWire/PulseAudio) and Android, making it useful for watching foreign-language videos, calls or streams without separate subtitle files.

把电脑或麦克风声音转成实时翻译字幕。Live translated subtitles from computer or microphone audio.

anomalyco/hex

Push-to-talk dictation that stays on your Mac, now spreading to Linux and Windows — HEX is a native Rust dictation app: hold a shortcut, speak, release, and the transcription pastes into whatever app you're using. Transcription runs on-device via a local speech model, with optional provider-backed voice commands kept separate. Originally macOS-only, it now ships Windows and Linux betas, making it interesting for privacy-conscious users across platforms.

Local-first voice dictation for macOS. Hold a shortcut, speak, and paste into any app. Native Rust, on-device speech recognition, and a Linux beta.

opennookorg/betterwispr

Free, open-source, local-first dictation app for macOS. Hold a shortcut, speak, and text is typed into any app. On-device speech-to-text with Whisper, Parakeet and Apple speech. Private, offline, open-source Wispr Flow alternative.

oboroge0/hayamimi

CPU-only multilingual live transcription that beats Whisper on Japanese CER — hayamimi routes each utterance to a specialist quantized ONNX model per language via sherpa-onnx instead of one general Whisper model, hitting 3.8% CER on Japanese broadcast audio vs Whisper-large-v3-turbo's 13.8%, at 10-50x realtime on a 6-core CPU, under 2GB RAM, with no GPU or cloud API.

早耳 - Real-time multilingual speech-to-text on CPU only. Live subtitles, browser dashboard, speaker labels, translation. No GPU, no cloud.

SaarthurR/atticus-mac

A voice assistant that runs your Mac with Claude Pro — Atticus is a free, open-source macOS voice assistant triggered by a hotkey. It uses your existing Claude Pro or Max plan instead of an API key, skips AI for repeated actions (0.08s app launches), and learns your shortcuts and personal context over time.

Fast, free voice assistant for your Mac. No API key: runs on your Claude Pro or Max plan. Sends background agents for big jobs, remembers you and learns your shortcuts. Hold ⌃⌥Space and talk.

nyosegawa/speech.cpp

A whisper.cpp-style toolkit covering TTS, ASR and VAD in one binary — Built on ggml like whisper.cpp, speech.cpp bundles text-to-speech (Qwen3-TTS, Irodori-TTS), speech recognition (Qwen3-ASR, ReazonSpeech, parakeet) and Silero VAD behind one C API and CLI. It runs on CPU, Metal or Vulkan, offers live mic transcription, and serves an OpenAI-compatible audio API, useful for building local voice assistants without cloud dependencies.

Speech synthesis, speech recognition and voice activity detection for voice conversation, in C++ on ggml: Qwen3-TTS, Irodori-TTS, Qwen3-ASR, ReazonSpeech, parakeet and Silero VAD on Metal, Vulkan and the CPU, with an OpenAI-compatible server and live transcription.