v1.12.5

vercel/streamdownv1.12.5Apr 8, 2026by FranckyB

AI Summary

This release integrates the Fish Speech S2 Pro TTS engine with extensive language support and introduces GPU kernel compilation for significantly improved generation performance.

Key Highlights

Fish Speech S2 Pro (4B parameter TTS model) added as a new engine supporting 80+ languages.
Triton/Inductor GPU kernel compilation with persistent caching for faster repeat generations.
Windows-specific patches for compilation compatibility and informative cache warnings.
TTS model manager methods added for model loading and reference audio encoding.

New Features

Fish Speech S2 Pro engine integration
Triton/Inductor GPU kernel compilation
Windows-specific compilation patches
80+ languages and 15,000+ inline expression tags support

Full Release Notes

Fish Speech S2 Pro Integration and Features:

Added Fish Speech S2 Pro (4B parameter TTS model) as a new engine in the UI, supporting 80+ languages and 15,000+ inline expression tags for fine-grained speech control. The model is auto-downloaded from HuggingFace, and [tag] syntax is supported and automatically stripped for other engines.

The TTS model manager (tts_manager.py) now provides get_fish_speech() and generate_voice_clone_fish_speech() methods, handling model loading, reference audio encoding, kernel cache status reporting, and batch/paragraph generation with all Fish Speech parameters exposed.

Performance and Stability Enhancements:

Implemented Triton/Inductor GPU kernel compilation with persistent caching in models/.cache for Fish Speech, significantly accelerating repeat generations. Added Windows-specific patches for compilation compatibility and informative first-run cache warnings.