🚀 Audio LLM Hidden Space Decoder

Pick an index manually or hit 'Pick Random Sample' to stream audio directly from EmoDB and compare all three execution layers.

Select Audio Index

0 534

Audio Playback (Streamed from HF)

Speaker ID

Ground Truth (Human Label)

1. Zero-Shot Text Generation Prediction

2. Global 80/20 Embedding Head Prediction

3. Leave-Speakers-Out Head Prediction

Global Embedding Classifier Class Probabilities