whisper-chess-tiny-de

whisper-chess-tiny-de

Fine-tuned Whisper-tiny for chess move recognition in Deutsch (German).

Part of the SpeakChess project — play chess by voice in EN / FR / DE / ES.

Performance

• Test WER: 0.01% on synthetic chess move evaluation set
• Domain: chess moves only (notation like Nf3, exd5, O-O)
• Optimized for browser inference via transformers.js (ONNX + INT8)

Files

• onnx/encoder_model_int8.onnx — INT8-quantized encoder (Conv layers kept FP32 for WASM compat)
• onnx/decoder_model_merged_int8.onnx — INT8 merged decoder
• Standard Whisper tokenizer/processor files

Total runtime download: ~62 MB.

Usage (transformers.js)

import { pipeline } from "@huggingface/transformers";

const transcriber = await pipeline(
  "automatic-speech-recognition",
  "atamano/whisper-chess-tiny-de",
  { dtype: { encoder_model: "int8", decoder_model_merged: "int8" } }
);

const result = await transcriber(audio, { language: "de", task: "transcribe" });

License — Important

This model is licensed under CC BY-NC-SA 4.0. Summary:

• ✅ Free to use for personal projects, research, education, and non-commercial demos
• ✅ Free to share and adapt with attribution and same-license derivatives
• ❌ Commercial use is NOT permitted without a separate license
• ❌ Including the model in commercial products, paid services, or competing voice-chess offerings requires explicit written permission

For commercial licensing inquiries: [email protected]

The full license text: https://creativecommons.org/licenses/by-nc-sa/4.0/

All rights reserved beyond the CC BY-NC-SA 4.0 grant.

Training

See github.com/atamano/speakchess for the full training pipeline.

• Base: openai/whisper-tiny (39M params, 98% trainable via full fine-tune)
• Method: Full fine-tuning + SpecAugment + suppress_tokens whitelist (chess vocab only)
• Data: synthetic audio via Edge TTS (multiple accents per language) + augmentation at runtime