EXPERIMENTAL IMAGE SYSTEMS · 003

IMAGE/

Make images on your own hardware. Three independent engines: procedural creation, local neural generation, and high-fidelity AI photo editing.

RAM NOT REPORTEDBROWSER RUNTIME

OUTPUT / 001READY TO GENERATE

NO RENDER YETEnter a prompt and generate an image.

ENGINESYNTH

SIZE768²

SEED—

TIME—

SESSION / HISTORY0 renders

Your generated frames appear here. Nothing is uploaded or stored on a server.

GENERATION

Synthesis controls

PROMPT SCENE DIRECTIVES

SYNTH interprets scene, materials, lighting and colors. It creates new procedural artwork, not photorealistic AI scenes.

SCENEASPECT RATIO

RESOLUTIONCOLOR SYSTEM

Detail / frequency0.65 Atmosphere / fog0.55 Texture / grain0.15 Lighting0.65

SEED

Ready. Image creation runs locally.

SESSION ACTIONS

IMPLEMENTATION / LIMITS

SYNTH is our custom deterministic WebGL2 procedural renderer; prompt words steer a limited scene grammar. It does not contain a trained text-to-image neural network. NEURAL loads the third-party SDXS model into your browser through ONNX Runtime Web and can generate pictures outside SYNTH's fixed scene grammar, but first-use downloads and GPU memory consumption are substantial. Neither engine submits your text to a remote inference server. Model weights are retrieved from Hugging Face only when you choose to load them.

EDIT / SOURCE FRAMEUPLOAD AN IMAGE

⌗

IMPORT A PHOTOGRAPH Choose from gallery, capture a photo, or drop an image here.

BLUE OVERLAY = REGION TO CHANGE NO SELECTION

BRUSH SIZE 50 EDGE FEATHER 24

EDIT HISTORYSOURCE ONLY

Edited versions will appear here. Images remain local until you choose a GPU connection.

REFERENCE IMAGE + TEXT

Photorealistic editing

QWEN 2511

EDIT INSTRUCTION FREE FORM

REGION MODE INFLUENCE

OUTPUT SIZE AI MIX 100%

MODEL CONNECTIONNOT CONNECTED

Choose RunPod Hub for on-demand GPU editing or use your own ComfyUI GPU. Your phone handles the selection brush and blends the edited pixels locally.

GPU BACKEND

YOUR PRIVATE GPU GATEWAY URL PRIVATE ACCESS TOKEN

Your RunPod API key stays on the private Cloudflare Worker, never in this website. The token above authenticates only you; it is kept in this tab's memory, not saved to GitHub or local storage. Setup instructions ↗

RUNPOD HUB · PAY PER EDIT

Upload a photo, optionally select an area, then connect your private RunPod gateway or ComfyUI server.

The app does not impose preset creative categories or watermarks. External server policies, model licenses, and applicable laws still apply. Masking preserves untouched pixels outside the selected region; seamless edits depend on the model and selected context.

QWEN MODEL AND WORKFLOWS ↗ · OFFICIAL TEMPLATE ↗