Image & video · DWG 03
ima2-gen
A local image and video studio for people and coding agents
ima2-gen is a local studio for generating, branching and cleaning up images and video on your own machine. Use GPT, Grok, Gemini, NovelAI and ComfyUI in one place — from the browser, the Mac app, the CLI, or through your coding agent.
Spec
- Install
npm install -g ima2-genima2 serve- Category
- Image & video
- License
- MIT
- Repository
- github.com/lidge-ai/ima2-gen
- GitHub · npm
- 780+ stars1,100+ weekly downloadsSeptember 25, 2026 as of
Features
- FIG. 01
Many models, one workbench
Add a prompt and references, pick a lane and model. Every result keeps its prompt and settings so you can continue from it.
- FIG. 02
Branch in a node graph
Push a good image in several directions at once. Each branch remembers its parent, so nothing gets overwritten.
- FIG. 03
Real transparency and cleanup
Annotate, erase and cut backgrounds on the canvas, then export with a real alpha channel.
- FIG. 04
A CLI agents can call
Coding agents call commands like
ima2 gen,ima2 editandima2 vectorizedirectly to produce website assets.
Use it when
- You want your coding agent to create the images a site or deck needs
- You want several versions of one shot to compare side by side
- You need transparent-background assets such as logos or icons
FAQ
Which accounts do I need?
A login or API key for the lane you use — for example ChatGPT OAuth or an xAI account.
Where are results stored?
Generated images and history stay in a local folder.
How do agents use it?
The package ships agent skill documents, so a coding agent can read the CLI usage and call it right away.