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
docs(README): benchmark shows flagship Fun-ASR-Nano vLLM speed (340x), not the torch number (#3023)
11fe7db zhifu gao committed 5 months ago
docs(README): de-duplicate Quick Start (remove repeated Fun-ASR-Nano section, fix double generate + model flags) (#3021)
d076eeb zhifu gao committed 5 months ago
docs(model_selection): lead with flagship Fun-ASR-Nano (#3020)
2ad0514 zhifu gao committed 5 months ago
docs(README): lead Quick Start with the flagship Fun-ASR-Nano (SenseVoice kept as the CPU/multilingual option) (#3019)
0b01332 zhifu gao committed 5 months ago
fix(glm_asr): dedup result keys for duplicate audio basenames (#3014)
bdf7cc6 Leoy committed 5 months ago
docs(README): use canonical iic/ ModelScope namespace (was legacy damo/) (#3017)
384a556 zhifu gao committed 5 months ago
docs(README): fix broken quickstart (missing input=), remove internal growth-plan link, drop empty docker block, align vLLM speedup (#3016)
6ee6509 zhifu gao committed 5 months ago
docs(README): update What's New with June updates (llama.cpp/GGUF runtime, q8, releases) (#3015)
bda08ae zhifu gao committed 5 months ago
feat: q8_0 export option (--wtype q8_0) for SenseVoice/Paraformer/Nano encoder GGUF (#3013)
9f1d220 zhifu gao committed 5 months ago
docs(README): promote llama.cpp/GGUF runtime to a first-class Deploy section (#3012)
631f44a zhifu gao committed 5 months ago
docs(vllm_guide): explain EmbedsPrompt/streaming partial/SPK/concurrency setting (#3011)
b201dfa qiulang committed 5 months ago
docs(vllm_guide): let vLLM pin torch/torchaudio in the installation steps (#3009)
e192f0c qiulang committed 5 months ago
docs(README): make the quickstart runnable (add missing model.generate() call + CPU note) (#3010)
b0ef72a zhifu gao committed 5 months ago
fix(qwen3-asr): map ISO/short language codes to qwen-asr canonical names (#3008)
ce7788c montvid committed 5 months ago
docs(vllm_guide): drop stale repetition_penalty hardcode note (#3007)
fa695e0 zhifu gao committed 5 months ago
fix(glm_asr): warn when vLLM dtype=fp16 (degraded output) (#2993)
d5ea9eb Leoy committed 5 months ago
fix: detok review findings (null gguf_get_arr_str guard + utf-8 tokens.json read) (#3005)
9cc0ebc zhifu gao committed 5 months ago
feat: print transcription text in the binaries (embed vocab, in-binary detok) (#3004)
807bcd5 zhifu gao committed 5 months ago
test: numerical regression harness (frozen golden vs ggml/VAD/CIF/CTC output) (#3003)
3610b75 zhifu gao committed 5 months ago
ci: drop macos-x64 (Intel runner unavailable; arm64 covers modern Macs)
6840f0f zhifu gao committed 5 months ago
fix: B1 script portability (Hugging Face CLI fallback + friendly missing-dep error) (#3002)
49bc9b5 zhifu gao committed 5 months ago
build: define NOMINMAX/_USE_MATH_DEFINES for MSVC (Windows portability)
7cdc280 zhifu gao committed 5 months ago
build: define NOMINMAX/_USE_MATH_DEFINES for MSVC (Windows portability)
d0f355f zhifu gao committed 5 months ago
ci: cross-platform prebuilt binary release for llama.cpp runtime (#3001)
b902479 zhifu gao committed 5 months ago
docs: build note — copy funasr-common alongside the example (shared miniaudio audio loader) (#2996)
4a93f29 zhifu gao committed 5 months ago