planning: Ichigo Encoder

[dan-homebrew]

Goal

• Paralinguistic emotions: emotions, i.e. acoustic encoder
• We are using a semantic encoder i.e. WhisperVQ, so acoustic tokens are lost
• Parallel tokens, WaveLM
• Need to have a target

[hahuyhoang411]

Good papers for this topic:

1. https://arxiv.org/abs/2402.05755
2. https://arxiv.org/abs/2402.12226