Blame for README.md

---
language:
  - ja
pipeline_tag: automatic-speech-recognition
base_model:
  - KeisukeMiyamoto/zipformer-cpt
  - KeisukeMiyamoto/lfm2.5-230m-cpt
datasets:
  - KeisukeMiyamoto/whisper-cluster-audio
  - KeisukeMiyamoto/audio-transcript
  - KeisukeMiyamoto/lambda-audio-courpus
metrics:
  - cer
tags:
  - japanese
  - speech-to-text
  - lambda-audio
  - zipformer
  - lfm2
  - pytorch
model-index:
  - name: LambdaAudio v6
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          type: google/fleurs
          name: FLEURS Japanese
          config: ja_jp
          split: test
        metrics:
          - type: cer
            name: CER (%)
            value: 6.05
---

# LambdaAudio v6

LambdaAudio v6 is a Japanese automatic speech recognition (ASR) model that converts speech into Japanese text. It combines a Zipformer audio encoder with an LFM2.5 language decoder, both adapted to Japanese audio or transcripts before joint training.

On Japanese subset of Google FLEURS ASR benchmark, it achieves a character error rate (CER) close to Whisper large-v3-turbo (**6.05% vs. 5.83%**) while transcribing audio **2.3× faster**.

## Model details

| Field | Value |
| --- | --- |
| Developer | Keisuke Miyamoto |
| Task / language | Speech-to-text / Japanese (`ja`) |
| Model type | Encoder-decoder with an audio adapter and gated cross-attention |
| Model size | 350M parameters |
| Audio encoder | [KeisukeMiyamoto/zipformer-cpt](https://huggingface.co/KeisukeMiyamoto/zipformer-cpt) |
| Text decoder / tokenizer | [KeisukeMiyamoto/lfm2.5-230m-cpt](https://huggingface.co/KeisukeMiyamoto/lfm2.5-230m-cpt) |
| Input | 16 kHz audio, up to 30 seconds per segment |
| Output | Japanese transcription text |
| Audio feature interval | 20 ms after the encoder; 40 ms after the adapter |

## Architecture

The Zipformer encoder extracts features from 16 kHz audio. An audio adapter reduces the sequence length and projects these features into the decoder's hidden dimension. Gated cross-attention in the LFM2 full-attention blocks lets the decoder use audio features to predict the next text token.

![LambdaAudio architecture: Zipformer encoder, audio adapter, and LFM2 decoder with gated cross-attention](images/lambda_audio_architecture.png)

## Training data

| Stage | Dataset | Sources and labels |
| --- | --- | --- |
| Encoder continued pretraining | [Whisper Cluster Audio](https://huggingface.co/datasets/KeisukeMiyamoto/whisper-cluster-audio) | Internet Archive, publicly available podcasts, and radio; 500 cluster labels derived from Whisper large-v3-turbo encoder features |
| Decoder continued pretraining | [Japanese Audio Transcripts](https://huggingface.co/datasets/KeisukeMiyamoto/audio-transcript) | Japanese transcripts from NHK archives and technology podcasts |
| Joint ASR training | [Lambda Audio Corpus](https://huggingface.co/datasets/KeisukeMiyamoto/lambda-audio-courpus) | NHK archives, technology podcasts, and ReazonSpeech |

## Benchmarks

Each cell shows **CER (%) / total transcription time (seconds)**. Lower values are better.

| Model | ReazonSpeech | Common Voice 25 Japanese | FLEURS Japanese |
| --- | ---: | ---: | ---: |
| **[LambdaAudio v6](https://huggingface.co/KeisukeMiyamoto/lambda-audio-v6)** | **10.75 / 121.55** | **23.87 / 167.84** | **6.05 / 20.75** |
| [Whisper large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) | 11.42 / 354.04 | 26.40 / 592.25 | 5.83 / 47.82 |
| [Kotoba-Whisper v2.2](https://huggingface.co/kotoba-tech/kotoba-whisper-v2.2) | 11.70 / 322.53 | 22.89 / 541.39 | 7.35 / 46.15 |
| [ReazonSpeech wav2vec2](https://huggingface.co/reazon-research/japanese-wav2vec2-base-rs35kh) | 14.01 / 49.35 | 31.43 / 68.22 | 15.04 / 9.22 |
| [Parakeet tdt_ctc-0.6b-ja](https://huggingface.co/nvidia/parakeet-tdt_ctc-0.6b-ja) | 10.62 / 55.10 | 19.99 / 82.28 | 5.94 / 9.74 |
| [Cohere Transcribe 03-2026](https://huggingface.co/CohereLabs/cohere-transcribe-03-2026) | 6.45 / 176.00 | 21.58 / 227.42 | 4.36 / 30.77 |

### ReazonSpeech

![CER and transcription time on ReazonSpeech test](images/reazonspeech_test.png)

### Common Voice 25 Japanese

![CER and transcription time on Common Voice 25 Japanese test](images/common_voice_25_ja_test.png)

### FLEURS Japanese

![CER and transcription time on FLEURS Japanese test](images/fleurs_ja_test.png)