---

Cubric Vision is a desktop workspace for generating images and video on your own machine. It runs ComfyUI as its engine — curated models, tuned workflows, no node graphs to wire up. You type a prompt, pick a model, and refine the result with masking, detailing, upscaling, and video tools. Your prompts, images, videos, and project files stay on your disk.

Free, open source, and made by Mad Pony Interactive. No accounts. No API fees. Runs on your machine — remote GPU optional.

What it does

• Image generation — a curated lineup of the best open-source image models, from fast photographic generators to instruction editors that change only what you ask. Each ships with a tuned workflow so you get strong results without parameter fiddling. Every release adds the latest models and workflows.
• Video generation — text-to-video and image-to-video in stages: preview first, then take the shot further only when it's worth the render time. The lineup tracks the best open-source video models, including ones that generate with sound.
• Remote GPU (optional) — can't run the heaviest models on your own machine? Rent a cloud GPU on demand and generate on it instead, billed to your own account. Everything stays local until you connect.
• Masking and detailing — brush masks or auto-detect, inpaint any region, refine faces and details, export masks.
• Editing and upscaling — instruction-based image editing, background removal, and model-based generative upscalers for images and video, plus custom upscaler support.
• Video tools — interpolate, resize, crop, upscale, combine, and export clips without leaving the app.
• Projects and history — every generation lands in a project with full history. Compare results side by side, branch from any earlier image, and find everything again in the gallery.
• LoRAs and custom models — point the app at your own LoRA and upscaler folders alongside the curated lineup.
• One-click setup — the app installs its own ComfyUI engine and downloads models for you. After install, no internet needed for local generation.

Video generation

Text-to-video and image-to-video from the latest open-source video models — runs locally, staged previews so you spend compute only on shots worth finishing. Some models generate with sound and support first- and last-frame guidance.

https://github.com/user-attachments/assets/dd4c46e8-6941-49a2-b279-949fa1815aae

Download and install

Grab the portable build for your platform from GitHub Releases — no installer, just extract and launch. On first run the app sets up its ComfyUI engine and asks where to store models.

Windows: extract the zip and run CubricVision.exe. Windows will show "Windows protected your PC" — click More info, then Run anyway. The builds are not code-signed, so this warning is expected.

macOS: downloaded builds are quarantined. Clear it, then launch:

xattr -dr com.apple.quarantine "<extracted folder>"

Then double-click start.command.

Linux: extract the tarball and run ./start.sh. Engine setup needs git — the installer will offer to install it if it's missing.

Step-by-step instructions are in the installation guide.

Every build is free and public on GitHub Releases. If Cubric Vision is useful to you, you can fund its development on Patreon (recurring) or Gumroad (one-off) — support, not a paywall.

What you'll need

Windows 10/11, Linux, or macOS 14 (Sonoma) or later on Apple Silicon — every Mac with an M-series chip can run macOS 14 or newer.

Each model shows its own memory needs in the app, and lighter tiers are available for smaller machines. If a model is too heavy for your GPU, you can run it on a rented remote GPU instead.

Documentation

The full user guide lives at docs.cubric.studio: getting started, projects, prompt box, image tools, video tools, models, gallery, history, and hotkeys.

Community and support

• Discord — questions, feedback, and build announcements.
• GitHub Issues — bug reports.
• GitHub Discussions — feature requests.
• Patreon — a recurring three-tier subscription with tutorial project files and a direct line to the developer. Funds ongoing development.
• Gumroad — a one-off donation if you'd rather support the build without a subscription.

Vision is the first app in the Cubric Studio family. Audio and Prompt are planned siblings — all local, all open source.

For developers

Want to run from source or contribute? Start with docs/DEVELOPMENT.md, then read CONTRIBUTING.md before opening a PR. For security-sensitive reports, see SECURITY.md.

License

Cubric Vision is licensed under AGPL-3.0-only. Portable builds ship with readable app source — open source isn't a marketing line here, it's how the app is distributed.