OOJOO Lab — local transcription Model: Qwen3-ASR-0.6B, Qwen team / Alibaba Cloud. Apache License 2.0. https://huggingface.co/Qwen/Qwen3-ASR-0.6B https://github.com/QwenLM/Qwen3-ASR Browser ONNX conversion: jiangzhuo9357/Qwen3-ASR-0.6B-ONNX. Revision: 4a01b95fafe2c9e3af77e33c18bbb7de349c62f6 https://huggingface.co/jiangzhuo9357/Qwen3-ASR-0.6B-ONNX Conversion provenance: andrewleech/qwen3-asr-onnx and the model card. Converted model weights remain under Apache License 2.0. Inference runtime: Microsoft ONNX Runtime Web 1.31.1, MIT. See onnxruntime-MIT.txt. The WASM binary is distributed losslessly compressed and decompressed on-device. Tokenizer: Hugging Face Tokenizers.js 0.2.0, Apache License 2.0. See tokenizers-Apache-2.0.txt (also the full Apache 2.0 license text for the model). FFT: fft.js 4.0.4, Fedor Indutny, MIT. See fft-MIT.txt. Download integrity: @noble/hashes 2.0.1, MIT. See noble-hashes-MIT.txt. OOJOO Lab's browser inference adapter and 400-point Bluestein STFT are independent implementations of the published ONNX input/output contract and the model's preprocessing configuration. No application code from the Sokuji AGPL benchmark harness is distributed with this page. Model/runtime preparation uses GET requests for public model and engine files. Audio, filenames, vocabulary hints and transcripts are not sent in requests.