Fish-Speech 1.5

Fish-Speech 1.5, from Fish Audio (originally OpenAudio), is an open text-to-speech model trained on over one million hours of multilingual audio across roughly 13 languages. Version 1.5.0 (December 2024) improved expressiveness, stability, and cross-language voice cloning, and the project accumulated 31,000+ GitHub stars on the strength of its emotional control.

The license split matters: code is Apache 2.0 but the weights are CC-BY-NC-SA-4.0 — non-commercial use only, with production use routed to the Fish Audio API. Fish Audio has since raised $52M and launched the S1/S2 generations; v1.5 remains the fully open checkpoint. Voice-cloning capability also demands careful consent practices.