ima2-gen
EN KO
GitHub ↗
Local AI studio · Image + Video

Branch the
images you
love.

An AI studio for the way you actually iterate. Generate images in Classic, fan out in Node, create videos with Grok, and clean up in Canvas. Free ChatGPT + SuperGrok is all you need.

$ npm install -g ima2-gen && ima2 serve
One-Click Install (no npm required)

Don't have Node.js? The script detects your environment, installs Node LTS if needed, then installs ima2-gen.

macOS curl -fsSL https://lidge-jun.github.io/ima2-gen/install-mac.sh | bash
Windows irm https://lidge-jun.github.io/ima2-gen/install-windows.ps1 | iex
Linux / WSL curl -fsSL https://lidge-jun.github.io/ima2-gen/install-linux.sh | bash
한국어 README →
ima2-gen video playback with gallery sidebar showing generated images and videos.

Two ways to make. Canvas to refine.

Classic for one strong frame or a multimode batch. Node for a tree of variations from a parent you locked in. Canvas for cleanup before the next prompt.

ima2-gen Classic mode with prompt composer, references, and a generated team portrait.
Classic

One prompt, one strong frame.

Write, attach up to five references, generate one result or several multimode slots. Iterate until it lands. Continue from any result.

ima2-gen Node mode with connected generated cards and per-node metadata.
Node

One frame, ten directions.

Lock a parent. Fan out children — palette, framing, copy. Compare side by side. Running branches recover by request ID.

Video, batch, import.

Grok Video with storyboard continuity and keyframe copy, multimode candidate slots, Canvas Mode cleanup, and prompt-library imports from files, GitHub folders, and curated sources.

ima2-gen video generation with progress percentage and in-flight queue.
Video

Text, image, or references to video.

Generate short videos from text, a single image, or up to 7 references. Storyboard mode keeps character and scene continuity across clips; First/Mid/Last buttons copy keyframes from results.

ima2-gen multimode sequence showing four generating slots and active job history.
Multimode

Four candidates from one prompt.

Run a sequence from Classic mode, watch each slot progress, cancel when needed, and continue from the strongest result.

ima2-gen Canvas Mode showing annotation marks, sticky note, zoom controls, and canvas toolbar.
Canvas Mode

Clean up the winning frame.

Pan around a zoomed image, annotate target areas, clean backgrounds, and export transparent or matte-backed canvas versions.

ima2-gen prompt import dialog for reviewing GitHub folders, curated sources, and searched prompt candidates before import.
Prompt import

Bring prompt packs in.

Use the import dialog to drop files, preview GitHub folders, refresh curated sources, and review searched candidates before adding selected prompts to the library.

A good frame is a starting point.

Locking the parent and fanning out children is how visual ideas actually develop. ima2-gen makes it the default.

ima2-gen Node mode canvas with a parent node and child branches showing palette, framing, and copy variants.
  1. 01
    Generate a parent. A frame you actually want to keep.
  2. 02
    Fan out children. Palette swap, tighter crop, alternate copy margin — three nodes, three runs.
  3. 03
    Compare without loss. The parent stays. Children recover by request ID even after a refresh.
ima2-gen Settings workspace showing GPT OAuth active and an API key configured but disabled.

Sign in. Stay local.

Start with your free ChatGPT login or a SuperGrok subscription. No API key needed. Generated images, videos, and sessions stay on your disk — not on someone else's.

If the settings page says Configured but disabled, that means an API key exists in env/config but the current workspace is using GPT OAuth. Switch provider to API or Grok when you want those paths.

User-data folder · MIT licensed · Inspect, sync, version — your call.

One command. Then iterate.

Install globally. Sign in with ChatGPT or Grok. Start creating.

$ npm install -g ima2-gen
$ ima2 setup
$ ima2 serve

One-Click Install (no npm required)

macOS curl -fsSL https://lidge-jun.github.io/ima2-gen/install-mac.sh | bash
Windows irm https://lidge-jun.github.io/ima2-gen/install-windows.ps1 | iex
Linux / WSL curl -fsSL https://lidge-jun.github.io/ima2-gen/install-linux.sh | bash
npm install Node ≥20 MIT
When should I use Node mode or Canvas Mode?

Use Node mode to compare several directions from a result you like. Use Canvas Mode when the frame is close but needs targeted cleanup, background work, annotation, or alpha/matte export.

Where do my images live?

Locally. Generated images and session logs live in your user-data folder. Inspect, sync, version — your call. Image files are not uploaded to a separate remote store.

Do I need an OpenAI API key?

No for the default GPT OAuth path. ima2-gen can also use an OpenAI API key or bundled Grok/progrok when you explicitly choose those providers.

See all FAQ →