Vison v0.1.0 — first public release

First public release of Vison: local image and video generation on your own GPU. No account, no credits, no prompts leaving your machine.

Download

Grab the Windows installer (Vison-Setup-*.exe) from the assets below. It is unsigned, so Win first run: click
"More info" then "Run

What's in it

• Text-to-image and text-to-video through a C++/Vulkan backend built on
  stable-diffusion.cpp, , AMD or Intel).
• Image models: SDXL Turbo, Z-Image Turbo, FLUX.1 Schnell, Qwen-Image. Video: Wan 2.1/2.2, HunyuanVideo 1.5, LTX 2.5 Distilled.
• Image-to-image by attaching a starting image.
• Image and video upscaling with ESRGAN models.
• Add your own models by pasting a Hugging Face link, including GGUF upscalers.
• Searchable history of everything you've generated.
• Automatic VRAM check, so models that won't fit are disabled instead of failing mid-download.
• Optional Google sign-in. The app is fully usable signed out.
• MIT licensed.

Known limitations

• Windows only for now; macOS and Linux are not packaged.
• Tested mainly on a 6 GB laptop GPU. Qwen-Image, Wan 2.2 A14B, HunyuanVideo 1.5 and Lnot been run tocompletion on real hardware.
• Early software from a single developer: expect rough edges, and please open an issue if something breaks.

Full notes in CHANGELOG.md.