Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
Commits on main
feat: B1 packaging — one-command download, standalone convert, CI-friendly CMake (#3000)
66358c5 zhifu gao committed 5 months ago
fix: address FSMN-VAD review findings (portability + robustness) (#2999)
2775666 zhifu gao committed 5 months ago
feat: built-in FSMN-VAD (--vad) for all three runtimes — single-binary segmentation (#2998)
e615ed0 zhifu gao committed 5 months ago
feat: accept any audio input (any rate/channels, wav/mp3/flac) via miniaudio (#2994)
a83e1d9 zhifu gao committed 5 months ago
docs: link the new llama.cpp / GGUF (CPU/edge) runtime from the Deploy section (#2991)
4ab43a4 zhifu gao committed 5 months ago
docs: fix README streaming example (runnable + actually streams) (#2987)
7e01e94 zhifu gao committed 5 months ago
docs: make README quickstart runnable and output truthful (#2986)
82a1245 zhifu gao committed 5 months ago
docs: python wss server now supports multiple concurrent clients (#2985)
dd9487b zhifu gao committed 5 months ago
fix(cli): correct fun-asr-nano model id (Fun-ASR-Nano-2512) (#2984)
79ac9e7 zhifu gao committed 5 months ago
fix(cli): use sentence_info "sentence" field for segment text (#2983)
3cdb661 zhifu gao committed 5 months ago
fix(cli): valid SRT timestamps + clear json duration fields (#2982)
6f8d161 zhifu gao committed 5 months ago
feat(fun_asr_nano): batched VAD-segment decoding (~1.75x faster) (#2979)
c39ff39 zhifu gao committed 5 months ago
fix(fun_asr_nano): warn when vLLM dtype=fp16 (degraded output) (#2980)
8107365 zhifu gao committed 5 months ago
fix(load): clear FileNotFoundError for missing audio path (#2981)
e6ab4c3 zhifu gao committed 5 months ago
fix(fun_asr_nano): make bf16/fp16 inference work instead of crashing (#2978)
faffb45 zhifu gao committed 5 months ago
fix: avoid CUDA crash from repetition_penalty in Fun-ASR-Nano vLLM prompt-embeds mode (#2948) (#2974)
b7b952d Leoy committed 5 months ago
feat(runtime): add sentence-level timestamps to WebSocket 2pass server (#2969)
a3fb72e psk-github committed 5 months ago
feat: add agent-friendly CLI (funasr audio.wav --output-format json)
b8351fd 游雁 committed 5 months ago
fix: handle None scheduler_conf in trainer, add scheduler to VAD finetune
b710ffd 游雁 committed 5 months ago