🎙️ Voice Command Classifier

Fine-tuned Wav2Vec2 Transformer — 92% Accuracy

Upload a 1-second audio clip of someone saying yes, no, stop, or go.

The model will predict which word was spoken and show confidence scores for each class.

Model: facebook/wav2vec2-base fine-tuned on Google Speech Commands v0.01
Dataset: 2,000 clips (500 per class)
Training: Transfer learning — only 197K of 94M parameters trained