Echolancer Stage 3 Base
This is a TTS model pretrained on the pre-tokenized Emilia dataset. Since there's no speaker conditioning, the speaker is random at inference. This model has 1.3B parameters and was trained from scratch on 8xA100s for ~2.5 days
The training objective was standard next-token prediction on concatenated text-audio tokens.
Code
For more information including a Colab notebook, see the repository.