datasets, language, license, metrics, pipeline_tag, library_name, widget, model-index

datasets	language	license	metrics	pipeline_tag	library_name	widget	model-index
bond005/taiga_speech_v2 bond005/podlodka_speech bond005/rulibrispeech	ru	apache-2.0	wer	automatic-speech-recognition	transformers	example_title src Нейронные сети - это хорошо! https://huggingface.co/bond005/whisper-large-v3-ru-podlodka/resolve/main/test_sound_ru.flac example_title src К сожалению, система распознавания речи не всегда стабильна, особенно в шумных условиях. https://huggingface.co/bond005/whisper-large-v3-ru-podlodka/resolve/main/test_sound_with_noise.wav example_title src Мимо театра мальчик ходил довольно часто — белое, со взбитыми сливками, здание-торт. https://huggingface.co/bond005/whisper-large-v3-ru-podlodka/resolve/main/anna_matveeva_test.wav	name results Whisper Large V3 Russian Podlodka by Ivan Bondarenko task dataset metrics type name automatic-speech-recognition Speech Recognition name type args Podlodka.io bond005/podlodka_speech ru type value name wer 20.91 WER (with punctuation and capital letters) type value name wer 10.987 WER (without punctuation) task dataset metrics type name automatic-speech-recognition Speech Recognition name type args Russian Librispeech bond005/rulibrispeech ru type value name wer 9.795 WER (without punctuation)

Whisper Large V3 Russian Podlodka

This repository contains a fine-tuned Whisper Large V3 model for Russian speech recognition. It serves as the core transcription component of the Pisets system, specifically optimized for long audio recordings such as lectures and interviews.

The model was presented in the paper Pisets: A Robust Speech Recognition System for Lectures and Interviews.

System Architecture

The Pisets system implements a three-component architecture to improve recognition accuracy while minimizing hallucinations:

1. Wav2Vec2: For primary recognition and segmentation.
2. Audio Spectrogram Transformer (AST): For filtering non-speech segments.
3. Whisper (this model): For the final high-quality transcription.

Implementation

The complete source code and instructions for using the system (including generation of SRT and DocX files) can be found in the GitHub repository:

GitHub: https://github.com/bond005/pisets

Citation

If you use this model or the Pisets system in your research, please cite:

@article{bondarenko2026pisets,
  title={Pisets: A Robust Speech Recognition System for Lectures and Interviews},
  author={Ivan Bondarenko},
  journal={arXiv preprint arXiv:2601.18415},
  year={2026}
}