The controls beside the prompt box change depending on which tool you picked from the menu — not every tool accepts every setting. This page explains the ones you are likely to meet.
The limits belong to the tool, not to the product. The prompt box shows its own count as you type — 2000 characters on some tools, 5000 on others — and the resolution choices differ too: Apex Video offers 480p to 4K, while a partner model may offer 1K, 2K and 4K. Read the panel in front of you rather than memorising a number.
That applies to reference material too. How many images, videos or audio clips a tool accepts is set by that tool — Apex Video takes far more than the image tools do — and the panel states its own limits beside each field. The one thing that is the same everywhere: the button that starts the work shows its cost in credits before you press it.
If you are new: leave all of them alone. The defaults are chosen to work. Get comfortable with prompts first; come back when a specific thing is wrong.
Resolution
The choices depend on the tool — 480p/720p/1080p/4K on Apex Video, 1K/2K/4K on some partner models.
Larger costs more and takes longer. Generate at a normal size and upscale the one you keep — cheaper than generating everything at 4K.
Size and shape
Whether the picture is square, portrait or landscape.
Pick the shape to suit where the picture is going — portrait for a phone screen, landscape for a banner, square for most social posts. Getting this right at the start beats cropping afterwards, because cropping throws away part of the picture.
Larger sizes cost more and take longer. Generate at a normal size, then upscale the one you keep — that is cheaper than generating everything large.
Composition follows shape. The same prompt gives a different arrangement in portrait than in landscape, because the subject has to fit differently.
How many
Generates several at once from the same prompt. Each costs the same as one, so four images costs four times one image.
Worth it. Because results vary, four gives you a choice; one gives you a verdict. Two to four is a reasonable habit while you are still refining a prompt.
Seed
The seed is the starting point of the randomness. Same prompt, same settings, same seed gives you the same image again. Change the seed and you get a different one.
Normally it is random, and you should leave it that way. It becomes useful when:
- You got something you like and want to try small prompt changes on that same composition — keep the seed, edit the prompt.
- You want to reproduce an earlier result exactly.
A seed is not transferable between models. The same seed on a different model gives an unrelated picture.
This is also what Regenerate does. Running one of your own library items again pins the seed it was actually produced with, so it reproduces that result rather than varying it. Clear the seed first if you want a variation instead.
Some models manage their own seeds and do not offer the setting at all. If there is no seed control, that is why.
Negative prompt
Where you list what you do not want: blurry, watermark, text, extra fingers.
Two rules. Keep it short — a long list dilutes it. And do not put things here that you have also described in the main prompt, because you are then asking for the same thing twice, in opposite directions.
Guidance or prompt strength
How literally the model follows your words.
- Low — loose interpretation, more invention, often more natural-looking
- High — sticks closely to your description, can look forced or over-saturated
The default sits in a sensible middle. If pictures feel over-cooked or unnaturally intense, lower it. If they keep ignoring something you clearly asked for, raise it a little — but check your prompt first, since an ignored detail is more often a prompt problem than a setting problem.
Steps or quality
How much work goes into each image. More steps means more time and more credits, with diminishing returns — the difference between low and medium is obvious, the difference between high and highest usually is not.
Leave it at the default. Raise it only for a final version you are keeping.
Strength (when starting from an image)
Only appears when you have supplied a starting picture. It sets how far the result is allowed to drift from it.
| Strength | Result |
|---|---|
| Low | very close to your original, small changes only |
| Medium | recognisably the same scene, new treatment |
| High | your image as loose inspiration |
If a style change is not taking, raise it. If your subject stops being recognisable, lower it. See starting from an image.
Advanced or Pro mode
A toggle that reveals additional controls. Everything behind it is optional — the panel works fully without ever opening it.
Turn it on when a guide tells you to, or when you already know which extra control you are looking for.
Presets and saved settings
Once a combination works, you can save it and apply it again later instead of setting everything up by hand. There may also be ready-made presets to start from.
Presets set the controls; they do not lock them. Apply one, then change whatever you like.
How to change settings without confusing yourself
One at a time, and keep the prompt fixed while you do it. If you change size, steps and guidance together and the result is worse, you have three suspects and no evidence.