ComfyUI engine, Cubric workflow
ComfyUI powers the backend while Cubric Vision handles the workflow UI, so creators get production tools without node-graph overhead.
AI made easy
The output leads. The app follows.
Cubric Vision gives artists a familiar desktop interface for local AI generation. Install models, generate images, refine details with masks, preview video motion in stages, and keep every file on your machine.
GenerationImage and video
PrivacyRuns offline
Learning CurveMinutes, not weeks
SourceOpen and free
Interface
A desktop workspace, not a prompt toy.
ComfyUI's engine, without the engine room. Generate, mask, stage video, install models, and manage projects from one window — sane defaults instead of parameter sliders, no Python, no node graph.
Features
Built for artists getting into AI.
No node graph required. Cubric Vision packages powerful local workflows behind controls that feel closer to a creative tool than a research project.
Familiar interface
A desktop app with project history, previews, controls, and media workflows that make sense to visual creators.
Curated local models
Curated lineup tuned for 8GB+ VRAM, so you can create without hunting model setups or hardware guides.
Stable on every update
Every ComfyUI update can break custom nodes and workflows. Cubric Vision ships portable, tested, and version-locked, so generations keep working release after release.
Mask and detail any area
Target faces, hands, backgrounds, and local regions for controlled refinements exactly where the image needs work.
Video in stages
Preview motion first, then commit to a final render so GPU time goes where it matters instead of discarded full outputs.
Free forever
Cubric Vision is open source. Patreon supports development, not a paywall around the core app.
Video combine
Select clips in order and combine them into a single output video without leaving your project workflow.
Custom upscalers and LoRAs
Use your own upscale models and LoRAs to push detail and style while staying inside Cubric Vision workflows.
Auto-masking and mask export
Use auto-masking plus manual refinement, then export masks as real files for Photoshop or other tools in your pipeline.
Photoreal
The setup, already done.
Photoreal image generation with the work already handled — the right checkpoints, LoRAs, samplers, and defaults dialed in for you. Unfiltered, and light enough to run on modest hardware. You prompt; the rest is taken care of.
Instruction editing
Edit by describing the change.
Say what you want changed and only that changes — the rest of the image stays intact. Combine several reference images in one prompt to place a character, face, or garment from one into another. No node graph, no manual masking to get started.
Workflow
One workflow, your file at the end.
Pick a model
Install and run local image and video models without manual setup.
Create
Generate images or clips from a prompt, reference, or existing project work.
Refine
Mask, detail, upscale, crop, or build on existing work.
Keep or take further
Save outputs to your project folder, ready to share, post, or take into Photoshop, DaVinci, or any tool you already use.
Video
Direct the motion, not just the prompt.
Generate video from a prompt or a single still, then guide the motion with first-frame, last-frame, or single-frame control. Preview the movement in stages before committing GPU time to a final render — then trim, reverse, interpolate, upscale, and pull stills, all in the same project.
Anime & stylized
Anime that looks made, not generated.
Sharp lines, clean shading, strong character work — anime, illustration, and bold stylized looks by default. Fast even on modest GPUs, with a detailing pass that closes the gap on sharpness and text. Many distinct styles, one workflow.
Download
Get Cubric Vision.
Latest portable build lives on GitHub. Unzip it, run the launcher. No installer, no account, no telemetry.
Support the project
Keep Cubric independent.
Patreon support helps keep Cubric apps free and funds the wider Studio hub. The Patreon and Discord community are run under MadPonyInteractive, the studio behind Cubric.
£1 / month
- Supporter Discord role (custom colour + badge)
- Name credited in Cubric apps About panel
- Behind-the-scenes dev posts before YouTube
£5 / month
- Supporter Discord role (custom colour + badge)
- Name credited in Cubric apps About panel
- Behind-the-scenes dev posts before YouTube
£20 / month
- Supporter Discord role (custom colour + badge)
- Name credited in Cubric apps About panel
- Behind-the-scenes dev posts before YouTube
Every tier does exactly the same thing. You just choose how much to give. Nothing is paywalled, gated, or delayed — every Cubric Vision version ships to everyone on GitHub at the same time.
Not into monthly? Give once instead — Cubric Vision is pay-what-you-want on Gumroad, including free.
Give once on Gumroad
FAQ
Questions creators ask before installing.
Can I use my own models?
Cubric Vision uses a curated lineup selected from the open ecosystem. Workflows and models are fine tuned to deliver stronger results with less parameter fiddling.
Are there any hidden costs?
No. Cubric Studio apps are free and will stay free. Patreon funds the build, it does not gate it — every version ships to everyone on GitHub at the same time.
Is it cross-compatible?
Yes, it is built for cross-platform use. Early testing has been mainly on Windows, so first releases on other platforms may have bugs. Built-in error dialogs can submit reports directly to GitHub Issues, and those reports help us improve quickly.
Can I use my creations commercially?
Yes. Creations you make with Cubric apps are yours to use commercially.
Can I use the code to build my own commercial app?
No. Code you use or fork from the Cubric repositories must remain in open-source apps, per the project license.
Do I need internet to use Cubric Vision after install?
After engine and model install, no internet is required for local inference. We plan optional cloud execution paths for lower-end machines in the future. If you choose those services, then internet is required.
Where are my projects and outputs stored?
If you do not choose a custom path, projects are stored in your Documents folder. Generated media is saved into each project together with project-level settings you changed.
What GPU and VRAM do I need for image vs video workflows?
Lighter image models run on about 8GB VRAM with roughly 16 to 32GB system RAM. Larger, newer image models are more comfortable around 12GB VRAM with at least 32GB system RAM. Video workflows usually need about 12 to 16GB VRAM and roughly 32 to 64GB system RAM.
Can I move a project between two computers?
Yes. Drag and drop a project folder onto the Projects page and Cubric Vision loads the assets and project data. To reproduce the same generation settings, install the same models on the target machine.
Can I run Cubric Vision alongside Photoshop or DaVinci on the same machine?
Yes. Use the release-memory action in Cubric Vision to free VRAM when needed, especially on lower-VRAM GPUs. You can export assets to external folders, edit in other apps, then re-import and continue.
How do I report bugs or request a model or workflow addition?
When an error occurs, you usually get a popup that can send a report to GitHub. Please use it when possible. For model and workflow requests, use the Discord community.
Does Cubric Vision upload prompts or media to any cloud service?
By default, no. Prompts, images, and videos stay local. A future optional RunPod path is planned, where processing runs on infrastructure you control and rent. If you opt in to that mode, cloud transfer is expected.
Can I roll back to an older version if an update breaks my workflow?
Yes. You can install a previous release. Because models can live in a separate models folder, rolling back can avoid re-downloading large model files.