OpenASR/funasr-nano

<div align="center">

Fun-ASR-Nano · OpenASR

Compact bilingual speech recognition — Mandarin and English in an ~0.8B SAN-M + Qwen3 pack

   

Native speech-to-text in the [OpenASR](https://github.com/QuintinShaw/openasr) runtime — engineered for peak performance on CPU & GPU, no Python at inference time.

</div>

✨ Highlights

• —📊 Verified accuracy: 3.03% CER (Chinese), 2.57% WER (English) — benchmarked by OpenASR on frozen evaluation datasets, not upstream-reported numbers
• —🌐 Fixed bilingual Mandarin + English — stock Qwen3 BPE vocab, no language-selection prompt needed
• —📦 Three quantization tiers: fp16 / q8_0 / q4_k — delivered in OpenASR's native .oasr format; encoder floored at Q80 on the q4k tier
• —🦀 Native in OpenASR — .oasr packs run with no Python at inference, engineered for peak performance on CPU & GPU

🚀 Quickstart

# 1. Install the OpenASR CLI  ·  https://openasr.org
# 2. Pull a build (pick a quant — see the table below)
openasr pull funasr-nano:q8

# 3. Transcribe
openasr transcribe audio.wav --model funasr-nano

All builds for this model:

openasr pull funasr-nano:fp16
openasr pull funasr-nano:q8
openasr pull funasr-nano:q4

📦 Available builds

_{RTF = real-time factor on the fixed 11s JFK clip (lower is faster); RAM peak measured per pack in an isolated subprocess. JFK ΔWER compares each quantized build's JFK transcript to this model's fp16 JFK transcript, so it measures quantization drift rather than absolute recognition accuracy. q8_0 is the recommended default — near-reference quality at a fraction of the footprint.}

🧠 About Fun-ASR-Nano

Fun-ASR-Nano is an ~0.8B-parameter speech-recognition model from FunAudioLLM (Fun-ASR-Nano-2512). Its architecture pairs a FunASR SAN-M/DFSMN audio encoder with a 2-layer transformer adaptor feeding a stock Qwen3-0.6B decoder. Language coverage is a fixed Mandarin + English set. OpenASR distributes this model in three quantization tiers -- fp16, q80, and q4k -- packaged in the native .oasr runtime format for local inference.

⚙️ How these packs were made

Converted from FunAudioLLM/Fun-ASR-Nano-2512 with the OpenASR importer:

openasr model-pack import funasr-nano <src> <out>.oasr \
  --package-id funasr-nano --quantization {fp16,q8-0,q4-k}

The .oasr container is GGUF-backed; packs use zero-copy mmap weight binding and graph buffer reuse to keep peak memory low.

⚖️ License

These packs inherit the upstream model's license: Apache-2.0 (source). OpenASR packaging retains the upstream copyright and NOTICE; the only modifications are format conversion and quantization.

🙏 Acknowledgements

This pack is a redistribution of Fun-ASR-Nano-2512, created and released by FunAudioLLM (FunAudioLLM/Fun-ASR-Nano-2512) under the Apache License 2.0. OpenASR performs format conversion, quantization, runtime validation, and local-inference adaptation only; all model weights and training are the work of the original authors.

🔗 Links

• —🦀 OpenASR — <https://github.com/QuintinShaw/openasr>
• —🌐 Website — <https://openasr.org>
• —🤗 Upstream model — FunAudioLLM/Fun-ASR-Nano-2512