Fine-tune Orpheus 3B for speech generation
A documented training recipe for adapting an Orpheus text-to-speech checkpoint to a transcribed audio dataset.
Published by Axolotl
How to use it
The publisher walks through audio preprocessing and conversion to the format expected by the Orpheus checkpoint. The included configuration requires you to replace placeholders with your prepared dataset and output settings.
Use consented or appropriately licensed recordings and keep held-out text and audio for evaluation. After preprocessing, run the documented axolotl train finetune.yml command and compare pronunciation and intelligibility on the held-out set.
- Framework
- Axolotl
- Dataset
- User-supplied transcribed audio, prepared as documented.
- Hardware
- Publisher documents a NVIDIA GPU with at least 45 GB VRAM for the supplied settings.
- Access
- Public publisher guide and configuration; bring your own dataset.