one prompt box
GPT Image 2, Nano Banana Pro, FLUX.2 [pro], Seedream and the rest sit behind a single prompt box. Switching model does not mean switching tool — the prompt stays where it is and the form beneath it becomes the parameters that model actually accepts.
Free to start · no card required
gives you
One box, every model
Change the model and keep your prompt. The parameter form redraws itself underneath — no re-entering the brief to try a second opinion.
Controls read from the schema
Aspect ratio, resolution, steps, guidance, seed: each field comes from the model's own definition, so nothing is offered that the endpoint would only reject.
Start from a picture, not a paragraph
Models with an edit endpoint take a reference image and work from it — hold a subject steady while you change the light, the setting or the framing around it.
Every output is an input
Send any result back into the box as a reference, push it into an App to finish it, or drop it onto the canvas as a node. Nothing is a dead end.
Kept, searchable, reusable
Every run is stored with the prompt, the model and the settings that produced it. Find it again in the Gallery, group it into Collections.
A queue, not a spinner
Jobs report where they are as they move. You know whether to wait for this one or go and start the next.
one workspace
This list is read from the app itself, not written by hand — what is here is what is in the picker.
start to finish
Pick a model
Capability chips next to each one are derived from its schema — the resolutions and options listed are the ones it genuinely offers.
Write the prompt
Add a reference image if you have one. Open the advanced group if the shot needs a seed or a specific guidance value.
Generate and compare
Run the same brief through a second model without retyping it, and keep both. Comparison is the point of having them in one place.
Finish it in Apps
Relight it, re-angle it, cut the background out or upscale it for print — every App works on anything already in your gallery.
plainly
- Modes
- Text to image, and image to image where the model supports it
- Controls
- Per-model, generated from each model's own field schema
- Reference images
- Upload, or pick anything already in your library
- After the run
- Apps, Canvas, Collections, or straight back in as a reference
answered
All of them — the list on this page is read from the app itself rather than written by hand, so what you see here is exactly what is in the model picker today.
Yes. Models with an edit endpoint accept a reference image alongside the prompt, and the Apps are built for exactly this: they act on an image you bring, whether it was generated here or not.
They are kept, with the prompt, model and settings attached. The Gallery is searchable, Collections group things into boards, and any output can be reused as an input later.
That is decided by the model you ran, not by us — each one ships its own licence terms and we surface them next to the model. Most permit commercial use; check the specific model before a frame goes into a campaign.
No. It runs in the browser, and signing up gets you credits to try every model without a card up front.
with
Your next frame is one prompt away
Open the prompt box, pick a model, and see what comes back. No card, no setup, no onboarding call.