---
datasets:
- MrDragonFox/EN_Emilia_Yodas_616h
language:
- en
base_model:
- ibm-granite/granite-4.0-h-1b-base
pipeline_tag: text-to-speech
library_name: orpheus
---

# Granite TTS 1B

A hybrid (Mamba) TTS model based on IBM's Granite models. This is very much a proof of concept / experiment, so you probably shouldn't use it in production. Trained on only 600 hours of data to start.

The model is still training, I will be uploading checkpoints here. The model will be made public when training finishes.

Base model: [ibm-granite/granite-4.0-h-1b-base](https://huggingface.co/ibm-granite/granite-4.0-h-1b-base)
Dataset: [MrDragonFox/EN_Emilia_Yodas_616h](https://huggingface.co/datasets/MrDragonFox/EN_Emilia_Yodas_616h)
Architecture: Orpheus TTS
Framework: [VyvoTTS](https://github.com/Vyvo-Labs/VyvoTTS) (slightly modified)