OPENAI-COMPATIBLE / LOCAL-FIRST

Image2 Studio

Turn your gateway into a real image workflow.

Connect your Base URL, API key, and gpt-image-2. Plan in natural language, edit with references, run paid calls in a strict queue, and keep every version traceable.

MODEL
gpt-image-2
GATEWAY
OpenAI Compatible
QUEUE
Strict Serial
STORAGE
Local First

IMAGE2 OUTPUT / 001—008

See the work first.

Every image below is an original generated by Image2 for this redesign, spanning commercial photography, portraits, spaces, interfaces, games, and narrative work.

Original fashion portrait beneath a rain-soaked arcade
PORTRAIT / 01

Blue-hour arcade

Natural skin, rain reflections, and restrained future fashion.

Graphite wireless speaker product photograph
PRODUCT / 02

Graphite audio object

Materials, construction, and edge light that hold up close.

Chef serving a dish inside a contemporary noodle bar
EDITORIAL / 03

Midnight noodle bar

Human interaction and warm commercial food storytelling.

Dark creative planning application on a tablet and phone
INTERFACE / 04

Creative task console

Devices, hierarchy, and system states that feel shippable.

Black timber and glass studio above a cold ocean
SPACE / 05

Cold-coast studio

Buildable structure, weather, and warm interior light.

Original science-fiction engineer in an orbital maintenance bay
CHARACTER / 06

Orbital bay engineer

An original character, credible equipment, and an industrial world.

Glass whale floating through a night archive
ILLUSTRATION / 07

The glass whale archive

Paper, ink, and luminous material in a quiet surreal frame.

Four-frame film storyboard of a train arriving in mountain fog
STORYBOARD / 08

First train, four frames

One location, coherent continuity, and a consistent camera language.

DESKTOP SYSTEM

Not another prompt box.

Image2 puts conversations, references, generation controls, serial tasks, and versioned results on one quiet desktop surface. Your key stays in the OS credential store; images and conversations stay local.

Image2 Studio dark desktop workspace
LIVE INTERFACEConversation / References / Serial Queue / Versions
01 / CONNECT

Use your gateway

Enter an HTTPS Base URL and API key, choose the agent protocol, and set the agent and gpt-image-2 model names.

02 / COMPOSE

Describe and reference

Assign identity, product, pose, composition, material, and style roles to references.

03 / CONTROL

Run a strict queue

Every direction is its own task, with clear stop, resume, retry, and cost-control states.

04 / REFINE

Continue from any version

Mark regions for local changes, preserve parent-child history, and branch from older results.

TASK PIPELINE

Complex requests, executed in order.

The agent understands and plans. The desktop app turns every paid call into a visible task you can stop, resume, and inspect.

  1. INPUT

    Write the goal, add references

    One request can combine subjects, style, copy, ratio, and multiple reference images.

    01
  2. PLAN

    Split it into images

    Every output has a title, prompt, parameters, and explicit reference relationships.

    02
  3. RUN

    Execute strictly in series

    Only one image task runs at a time, avoiding gateway rate limits and surprise concurrency costs.

    03
  4. ITERATE

    Mark and continue

    Points, boxes, arrows, and painted regions compile into a structured edit request.

    04

LOCAL SECURITY

Gateway credentials do not belong in a web bundle.

API KEY
macOS Keychain / Windows Credential Manager
REQUEST
Only the configured OpenAI-compatible paths
HISTORY
Conversations, tasks, images, and annotations stay local
WEBSITE
The marketing site accepts no key and offers no online generation

GET THE DESKTOP BUILD

Your gateway. Your key.
Your image system.

Get the Windows or macOS build on GitHub, read the installation guide, or inspect the source.

Current installers are unsigned and may trigger an OS warning. Download only from this repository's Releases page.