How to use ImageGPT v3.6
Images
Upload between one and seven images. The first one is the image you edit: it sets the framing and the output proportions. The others are references: a face, an outfit, an object, a background, or a style.
In the request, refer to the images in upload order as @Image1, @Image2 … @Image7. The model
also understands plain words like "from the second photo," but tags are more reliable when there are
many images.
Masked editing
By default the model edits the whole image. To change only part of it, switch to By mask and
paint the area on @Image1. The model redraws only that area; everything else stays as it was in
the original.
- Paint a little beyond the object so the new part blends in with the light and edges around it.
- The other images still work as references: "Replace the teapot with the mug from @Image2".
- Standard or High sets how detailed the painted area is, not the size of the result.
Writing a request
Keep it short: one or two sentences, in English or Russian. The model expands the request into a detailed prompt on its own, so long descriptions are not needed.
- Describe the finished result, not the process: "she wears a yellow raincoat," not "try changing the coat."
- Say separately what must stay unchanged: the face, the pose, the background, the lighting.
- Put sign or poster text in double quotes, word for word, and keep it short.
- Skip words like "masterpiece," "8K," or "ultra-realistic": they make the image glossy and oversaturated.
Examples
Replace her black coat with a glossy bright yellow raincoat, keep everything else the same.
Dress the woman from @Image1 in the gown from @Image2. Keep her face, pose, and the background.
Add a wooden sign on the tiled wall that reads "PROBLEMBO CAFE".
Make this scene night-time: turn on warm light in the windows and street lamps, keep the
architecture unchanged.
Seat the people from @Image2–@Image7 at the counter next to the man, keep the diner and the night
lighting.
Turn this illustration into a realistic photograph.
NSFW style
The NSFW style helps the model with explicit scenes: without it, the model often softens such requests. Turn it on only for explicit edits. On regular edits the style may change faces and remove clothing nobody asked to remove, so keep "No style" for them. The style does not change the price or the processing time.
Mode
- Standard is the default mode: about 4 megapixels, faster and cheaper. It suits most edits.
- High renders about 8 megapixels. It draws fine details, hands, and people in the background better and takes about twice as long.
The result keeps the proportions of the first image, so there is no aspect ratio to choose. A single run can produce up to two variants.
Limitations
- Small or long text may come out with mistakes; short quoted text renders well.
- Hands and small figures in the background are sometimes distorted; this happens less often in High mode.