Todos os artigos

Generating and editing images

Two things catch people out here. The settings under the prompt box belong to the model, not to Sunra, so changing the model changes what you can set. And there is no separate edit mode: attaching a picture to an image generation is how you edit it.

Make an image from a prompt

Open a project, set the mode pill at the left of the prompt box to Image Gen, and the box asks you to Describe an image….

Three controls sit in the row underneath:

  1. The model button — a logo, not a word. Click it for the picker; each row is a model, with a tick on the one you are using. Choosing a model explains what differs between them.
  2. The settings pill, marked with sliders. Its label is a summary of the current run, and clicking it opens the full set — the sections described below.
  3. Generate, which carries the credit cost of exactly the run you have set up.

Type the subject, the setting, the light and the framing in plain words; style belongs in the prompt too, not in a separate field. Enter inserts a new line, so nothing is spent until you press Generate.

Edit a picture you already have

On a model that can take reference images, a row of tiles appears above the prompt with an Add reference image slot. Fill it from the + menu's Upload files, by dragging a file in from the Assets grid, or with Use as reference on any result you already have. Then describe the change rather than the whole picture — "put her in a red coat", not the original prompt again.

How many reference slots you get is the model's own limit, and a model that takes none shows no slots at all. Once a file is attached you can point at it from the prompt text with @ — @Image1, @Image2 — which is how you say which reference a sentence is about.

Nota
Editing endpoints usually take their dimensions from the picture you gave them. When they do, the Size section says Follows reference image instead of offering ratios — that is the model refusing to resize, not a control that failed to load.

What the settings do

The settings popover only shows the sections the chosen model actually has:

SectionWhat it changes
SizeThe aspect ratio — square, landscape, portrait and whatever else the model offers.
ResolutionHow large the output is, where the model has more than one option.
QualityOn the models that have tiers, how hard the model works. Higher tiers cost more credits.
ReasoningLow, Medium or High: how deeply the model thinks about your prompt. The same price at every level.
Output formatPNG, or SVG on the few models that draw vectors. SVG takes no reference images and one image per request.
ImageHow many images this run makes.

Changing any of them updates the number on the Generate button, so you can price a setting before you commit to it. Credits only leave your balance when you press it; a run that fails or is cancelled is refunded. See Credits and plans.

How many images you get

The Image section counts finished images, not requests. Pick four and you are charged for four.

Some models hand back a fixed batch per call — two or four at a time — so the choices step in those units and a note under the grid tells you how many runs your number works out to. The cost on the button is for the whole batch.

Generating several at once is the cheapest way to learn a model: one prompt, four results, and you can see which direction the model leans before you spend anything on refining.

Why a control you expected is missing

Nothing in this row is hidden behind a setting — if a section is absent, the model does not have it. The same goes for references: attach one to a model that cannot read images and Sunra says so rather than quietly dropping it.

So when an option matters to you, pick the model for it. Switch models and the row rebuilds itself around the new one, keeping your prompt.

What to do with the result

Finished images land in the project — in the chat or on the canvas, and always in its Assets tab. Hover a result, or open it, and you can:

  • Download it, as it came or in another file type.
  • Use as reference — it becomes the input for the next image, or the start frame for a video.
  • Reuse settings — loads that generation's model and settings back into the prompt box, so you can change one word and run it again.
  • Like it, or Save to Library into one of your own folders.
  • Drag it onto the Design page, the Timeline, or a node on Flow.

Nothing has to be exported and re-uploaded to go from one step to the next: a finished image goes straight back into the prompt box as the reference for the next change.

Next: Generating video turns a still into motion, and Projects and assets covers where every file is kept.

Artigos relacionados