Generate with two new video models, create vocal songs, review source transcripts before lip sync, and run Hermes inside Desktop.
Web
September 4, 2026
No update required
Desktop
0.3.12
Minimum version
Agent Plugin
0.2.26
Minimum version
MiniMax H3 adds text, frames, references, and 2K options; H3 Max provides a higher-quality frame-led path with model-specific input limits.
Version requirement: Desktop 0.3.5 or later
Lip-sync video translation now presents an editable source transcript before submission, while audio-only translation can still proceed directly.
Version requirement: Desktop 0.3.4 or later
Music requests can create a structured vocal song with lyrics and singing direction, or continue through the existing instrumental path.
Version requirement: Desktop 0.3.8 or later
Desktop detects Hermes, guides sign-in, configures the current project's MCP connection, and restores the session and selected model after restart.
Version requirement: Desktop 0.3.4 or later
Web and Desktop add Japanese and Traditional Chinese interface catalogs alongside English, Spanish, and Simplified Chinese.
Version requirement: Desktop 0.3.5 or later