v1.3.13

What's Changed

• docs(vllm_guide): explain EmbedsPrompt/streaming partial/SPK/concurre… by @qiulang in #3011
• docs(README): make the llama.cpp/GGUF (CPU/edge) runtime a first-class section by @LauraGPT in #3012
• feat: q8_0 GGUF export — half-size encoders, CER unchanged by @LauraGPT in #3013
• docs(README): refresh What's New with June updates by @LauraGPT in #3015
• docs(README): holistic fixes — broken quickstart, stray links, consistency by @LauraGPT in #3016
• docs(README): canonical iic/ ModelScope namespace by @LauraGPT in #3017
• fix(glm_asr): dedup result keys for duplicate audio basenames by @SuperMarioYL in #3014

Full Changelog: v1.3.12...v1.3.13

v1.3.12

What's Changed

• docs(vllm_guide): drop stale repetition_penalty hardcode note by @LauraGPT in #3007
• fix(qwen3-asr): map ISO/short language codes to qwen-asr canonical names by @montvid in #3008
• docs(README): make the quickstart runnable (missing model.generate call) by @LauraGPT in #3010
• docs(vllm_guide): let vLLM pin torch/torchaudio in the installation steps by @qiulang in #3009

New Contributors

• @montvid made their first contribution in #3008
• @qiulang made their first contribution in #3009

Full Changelog: v1.3.11...v1.3.12

FunASR llama.cpp runtime runtime-llamacpp-v0.1.2

Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime — SenseVoice, Paraformer and Fun-ASR-Nano with built-in FSMN-VAD (a whisper.cpp-style on-device ASR, strong on Chinese). Get a model with bash download-funasr-model.sh <sensevoice|paraformer|nano>, then run llama-funasr-cli / llama-funasr-sensevoice / llama-funasr-paraformer. No Python, no build. Docs: runtime/llama.cpp/README.md

v1.3.11

What's Changed

• docs: python wss server now supports multiple concurrent clients by @LauraGPT in #2985
• docs: make README quickstart runnable and output truthful by @LauraGPT in #2986
• docs: fix README streaming example (runnable + actually streams) by @LauraGPT in #2987
• Add llama.cpp / GGUF runtime (Fun-ASR-Nano, SenseVoice, Paraformer) by @LauraGPT in #2988
• docs: link llama.cpp / GGUF (CPU/edge) runtime from Deploy section by @LauraGPT in #2991
• ci: auto-create GitHub Release on version tag push by @LauraGPT in #2995
• docs: CPU benchmark vs whisper.cpp (Chinese ASR) by @LauraGPT in #2992
• feat: accept any audio input (any rate/channels, wav/mp3/flac) via miniaudio by @LauraGPT in #2994
• feat: built-in FSMN-VAD (--vad) — single-binary speech segmentation, no Python at runtime by @LauraGPT in #2998
• fix: FSMN-VAD review findings (MSVC M_PI, short-audio guard, tensor validation) by @LauraGPT in #2999
• feat: B1 packaging — one-command download, standalone convert, CI-friendly CMake by @LauraGPT in #3000
• docs: build note for funasr-common (A1 follow-up) by @LauraGPT in #2996
• ci: cross-platform prebuilt binaries for the llama.cpp runtime by @LauraGPT in #3001
• fix: B1 script portability (HF CLI fallback + friendly missing-dep error) by @LauraGPT in #3002
• test: numerical regression harness (frozen golden vs ggml/VAD/CIF/CTC output) by @LauraGPT in #3003
• feat: print transcription text in the binaries (in-binary detok) by @LauraGPT in #3004
• fix: detok review findings (null vocab guard + utf-8 tokens read) by @LauraGPT in #3005
• fix(glm_asr): warn when vLLM dtype=fp16 (degraded output) by @SuperMarioYL in #2993
• fix(glm_asr): honor sampling params in vLLM generate() by @SuperMarioYL in #2997

Full Changelog: v1.3.10...v1.3.11

FunASR llama.cpp runtime runtime-llamacpp-v0.1.1

Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime — SenseVoice, Paraformer and Fun-ASR-Nano with built-in FSMN-VAD (a whisper.cpp-style on-device ASR, strong on Chinese). Get a model with bash download-funasr-model.sh <sensevoice|paraformer|nano>, then run llama-funasr-cli / llama-funasr-sensevoice / llama-funasr-paraformer. No Python, no build. Docs: runtime/llama.cpp/README.md

FunASR llama.cpp runtime runtime-llamacpp-v0.1.0

Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime — SenseVoice, Paraformer and Fun-ASR-Nano with built-in FSMN-VAD (a whisper.cpp-style on-device ASR, strong on Chinese). Get a model with bash download-funasr-model.sh <sensevoice|paraformer|nano>, then run llama-funasr-cli / llama-funasr-sensevoice / llama-funasr-paraformer. No Python, no build. Docs: runtime/llama.cpp/README.md

v1.3.10

FunASR v1.3.10

New features

• Agent-friendly CLI: funasr audio.wav --output-format json for structured output
• Fun-ASR-Nano: batched VAD-segment decoding (~1.75× faster) (#2979)
• WebSocket 2-pass server: sentence-level timestamps
• serve_vllm.py: new --vad-model / --spk-model flags

Fixes

• Fun-ASR-Nano: bf16/fp16 inference no longer crashes; warn on degraded fp16 (#2980)
• Fun-ASR-Nano vLLM: fix CUDA crash from repetition_penalty
• CLI: valid SRT timestamps + correct JSON durations (#2982); use sentence_info text (#2983); correct model id Fun-ASR-Nano-2512 (#2984)
• Clearer error for missing audio path (#2981); respect explicit VAD silence threshold; handle None encoder/scheduler configs

Docs

• New CLI reference; clearer vLLM install guidance

Full changelog: v1.3.9...v1.3.10

v1.3.9: Wheel packaging + SenseVoice speaker diarization fix

What's New

Wheel packaging (fixes #2943)

FunASR now publishes a py3-none-any wheel alongside the source distribution. Installation is faster since pip no longer needs to build from source.

Bug fixes

• SenseVoice + speaker diarization: Fixed crash when using spk_model="cam++" with SenseVoice (auto-falls back to VAD-segment mode since SenseVoice doesn't produce word-level timestamps)
• torchaudio >= 2.11 compatibility: Added soundfile as intermediate fallback for users with newer torchaudio versions that removed legacy backends

Install / Upgrade

pip install --upgrade funasr

Full changelog: v1.3.3...v1.3.9

v1.3.3: Agent Integration — OpenAI API + MCP Server + funasr-server CLI

Highlights

This release makes FunASR a drop-in speech backend for AI agents.

New: funasr-server CLI

pip install funasr fastapi uvicorn python-multipart
funasr-server --device cuda

One command starts an OpenAI-compatible /v1/audio/transcriptions endpoint.

New: MCP Server

AI assistants (Claude, Cursor, Windsurf) can now transcribe audio directly.

New: OpenAI-Compatible API

Works with any agent framework: LangChain, AutoGen, CrewAI, Dify, Flowise, Open WebUI.

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="x")
result = client.audio.transcriptions.create(model="sensevoice", file=open("a.wav","rb"))

Bug Fixes

• Fixed hub="hf" parameter propagation to sub-models (v1.3.2)
• Fixed Qwen3-ASR ImportError masking

Upgrade

pip install --upgrade funasr

Links

• Agent Integration Guide
• OpenAI API Docs
• MCP Server Docs
• Benchmark

v1.3.2: HuggingFace Hub Fix + Performance Benchmark

What's New

Bug Fix

• Fixed hub parameter propagation — When using hub="hf", the parameter is now correctly forwarded to VAD/PUNC/SPK sub-models. Previously, users on HuggingFace would get 404 errors for sub-models. (#2859)

Improvements

• Updated PyPI metadata with better description, keywords, and project URLs
• Added comprehensive benchmark page: https://modelscope.github.io/FunASR/benchmark.html

Benchmark Results (PyTorch, GPU)

Model	Type	Speed
SenseVoice-Small	NAR	170x realtime
Paraformer-Large	NAR	120x realtime
Whisper-large-v3-turbo	AR	46x realtime
Fun-ASR-Nano	LLM	17x realtime
Whisper-large-v3	AR	13.4x realtime

Install / Upgrade

pip install --upgrade funasr

Quick Start

from funasr import AutoModel
model = AutoModel(model="FunAudioLLM/SenseVoiceSmall", hub="hf", vad_model="funasr/fsmn-vad", device="cuda")
result = model.generate(input="audio.wav")