OpenCreator v3.2.3
What's New
- Desktop now checks the local Codex configuration on first launch. If an existing ChatGPT sign-in or API key setup is ready, you can reuse it without entering credentials again; otherwise, a guided setup lets you choose a model provider.
- Added local Codex image generation for Creator workflows, using the available Codex account instead of requiring a separate image-provider setup. Each request generates one image.
- Expanded video downloads to public videos from YouTube, Bilibili, X, TikTok, Instagram, Douyin, Facebook, Xiaohongshu, and Pinterest. Some sources may require platform cookies.
- Added a Codex-native YouTube Shorts workflow for stick figure videos, with improved narration, visuals, output formats, and publishing copy.
- Added Whisper.cpp
large-v3-turbo as an optional local transcription model.
- Improved video translation previews, subtitle styling controls, result downloads, and workspace layout. Fixed Chinese labels appearing in the English SRT file picker and dual-format output control.
- Unified Creator issue reporting and added actionable diagnostics in the Agent panel, so configuration, provider, and task failures are easier to understand and resolve.
- Improved Desktop startup, shutdown, and Runtime validation performance, and updated the multilingual documentation and community guides.
OpenCreator Desktop
KrillinAI Server
KrillinAI CLI
SHA-256 Checksums
The macOS installers are signed and notarized. The Windows installer is currently not Authenticode-signed; verify its SHA-256 checksum before installation.
Automatic Update Assets
ZIP archives, ZIP blockmaps, and latest YAML files are used by the automatic updater. For manual installation, download only the installer listed above.
Build manifests and debug configuration files are not included as public release assets.