HojoAI Open-Sources Qwen3-Based Multilingual ASR Model Covering 9 Languages
rohanpaul_ai · x · 2026-10-16
HojoAI has open-sourced Hojo-ASR-Multi-V1 on Hugging Face (Apache 2.0), a conversational speech recognition model built on the Qwen3 LLM decoder, using an Encoder-Adapter-LLM architecture with a customized multi-frame acoustic fusion design.
- Trained via multi-stage modular training plus reinforcement learning, it handles noisy conditions, informal pronunciation, and oral corrections well
- Supports German, French, Spanish, Portuguese, Japanese, Italian, Arabic, Korean and Russian
- Ships with a hojo-asr PyPI package — pip install and a few lines of Python to run inference
A commenter suggests a rigorous eval: run it against an existing ASR stack under identical audio conditions, publishing hardware, settings, ground truth and raw outputs to find exactly where the 3.54 WER breaks.
Related event: HojoAI Open-Sources Multilingual ASR Model Hojo-ASR with 3.54 Average WER(2 posts)→
More from Multimodal
- Pollo AI launches Recast video restyling, 60% off for launch — SimplyAnnisa · 2026-10-16
- Tripo hosts first world-building hackathon in SF with World Labs and Runway talks — mhdfaran · 2026-10-16
- Kling 4.0 boosts text-to-video with character consistency and lip sync — CurieuxExplorer · 2026-10-16
- 5.8M-param Speech Event Detector adds laughs, breaths and背景 sounds to Whisper transcripts — Ayush Kumar Bar · 2026-10-16
- Huawei's SUPRA-Tex turns one image into relightable 3D materials with WYSIWYG fidelity — Chen Yang · 2026-10-16
- Giant Michael Jackson AI fashion images show the power of extreme proportions — OVolosin82152 · 2026-10-16