pannous/tensorflow-speech-recognition

Implements end-to-end speech-to-text conversion by converting audio spectrograms through sequence-to-sequence models with LSTM and dense architectures, supporting real-time microphone input via PyAudio. The project includes modular training pipelines and toy classifiers for numbers and speakers, designed as a standalone Linux solution leveraging public datasets like OpenSLR. **Note: This is archived/educational code—the maintainers recommend using Mozilla DeepSpeech or OpenAI Whisper for production use.**

2,176 stars. No commits in the last 6 months.