TOOLORA / Local reading edition
F5-TTS
Generate speech from reference audio, its transcript and target text, with chunked inference, multi-voice sequences and fine-tuning.
Overview
F5-TTS conditions new speech on reference audio and its transcript. Command-line and Gradio workflows support chunked synthesis, multiple styles and speakers, plus training and fine-tuning. Omitting the reference transcript invokes an additional recognition model. Code uses MIT, while official pretrained weights use non-commercial CC-BY-NC terms; code, weights and training data have distinct usage conditions.
Official sources and references
- Official websitehttps://github.com/SWivid/F5-TTS
- Documentationhttps://huggingface.co/SWivid/F5-TTS
Descriptions summarize public information; consult official websites for features, pricing and regional availability.