img2img
img2img is the short label creators use for image-to-image workflows. This workspace is that loop — reference in, edited image out.
If you already know img2img from Stable Diffusion or other tools, the idea is familiar: strength of change comes from your prompt and model choice, anchored by the uploaded image.
Use img2img here when you want a hosted playground with credit-based models instead of running local weights. Pick Nano Banana for cheap drafts or GPT-4o Image when prompts must land precisely.
New users still get 6 free credits to test img2img before paying.
img2img tips if you came from local Stable Diffusion
This host does not expose a numeric denoise slider like every local UI. Translate your old habits as follows.
- Treat “low denoise” as a prompt that lists many keep-constraints (face, outfit, camera angle) and a cheap first model (Nano Banana, 3 credits).
- Treat “high denoise” as fewer keep-constraints and a clearer scene rewrite; expect more drift and budget a second run.
- Seed locking from desktop UIs is not the control to chase here — iterate with saved prompts and the same reference file instead.
- Multi-reference ControlNet-style stacks map best to Nano Banana’s higher reference limit (up to 10), not GPT-4o Image (1 file).
- Batch the same prompt across Flux Kontext (4 credits) only after Nano Banana proves the wording; do not burn 6-credit models on first guesses.
FAQ
Is img2img different from image to image?
No — img2img is the same category, just the shorter community name.
Do I need to install anything for img2img?
No install. Run img2img in the browser with an account and credits.
Which model should I pick first for img2img?
Start with Nano Banana to learn the loop cheaply, then try Flux Kontext or GPT-4o Image if you need different edit behavior.
Where is the denoise strength control for img2img?
There is no single denoise dial on this page. Encode strength in the prompt’s keep/change split and in which model you choose.
Can I bring LoRAs into this img2img workspace?
No custom LoRA upload here. You pick from the hosted image models listed in the playground.
Does img2img support inpainting masks?
This workspace is full-frame image-to-image from references and prompts. Masked inpaint is not exposed as a separate tool on this URL.