---
language:
  - zh
  - en
  - ja
  - yue
license: apache-2.0
library_name: transformers
pipeline_tag: automatic-speech-recognition
tags:
  - speech-recognition
  - asr
  - end-to-end
  - multilingual
  - streaming
---

<div align="center">

### ⭐ Powered by [FunASR](https://github.com/modelscope/FunASR) — please give us a GitHub Star!

This model is part of the **FunASR** ecosystem — one industrial-grade open-source toolkit for **ASR · VAD · punctuation · speaker diarization · emotion / event · LLM-ASR**. A Star really helps the project (and keeps you updated):

[**🌟 FunASR**](https://github.com/modelscope/FunASR)  ·  [**🌟 SenseVoice**](https://github.com/FunAudioLLM/SenseVoice)  ·  [**🌟 Fun-ASR**](https://github.com/FunAudioLLM/Fun-ASR)  ·  [**🌟 FunClip**](https://github.com/modelscope/FunClip)

</div>

# Fun-ASR-Nano (Hugging Face Transformers)

This is the Hugging Face Transformers-compatible version of [Fun-ASR-Nano-2512](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512).

Fun-ASR-Nano is an end-to-end speech recognition model by [FunAudioLLM](https://github.com/FunAudioLLM), trained on tens of millions of hours of real speech data. This checkpoint supports Chinese, English, and Japanese; its Chinese coverage includes 7 dialect groups and 26 regional accents. For 31-language recognition, use the separate [Fun-ASR-MLT-Nano-2512](https://huggingface.co/FunAudioLLM/Fun-ASR-MLT-Nano-2512) checkpoint.

## Transformers quickstart

Fun-ASR-Nano support is being added to Transformers in [huggingface/transformers#46180](https://github.com/huggingface/transformers/pull/46180). Until it is included in a Transformers release, use a build containing that PR.

```python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor

model_id = "FunAudioLLM/Fun-ASR-Nano-2512-hf"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForSpeechSeq2Seq.from_pretrained(
    model_id,
    dtype=torch.bfloat16,
    device_map="auto",
)

audio_url = "https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512/resolve/main/example/en.mp3"
inputs = processor.apply_transcription_request(audio=audio_url, return_tensors="pt").to(model.device)

generated_ids = model.generate(**inputs, max_new_tokens=200)
generated_ids = generated_ids[:, inputs.input_ids.shape[1] :]
print(processor.batch_decode(generated_ids, skip_special_tokens=True)[0])
```

Expected transcription:

```text
The tribal chieftain called for the boy, and presented him with fifty pieces of gold.
```

For batch inference, training, `torch.compile`, benchmarks, and the full model description, see the [Transformers documentation](https://github.com/huggingface/transformers/blob/main/docs/source/en/model_doc/fun_asr_nano.md) and the [original model card](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512).