Knowledge base · images
Reve 2.0
Generating and editing images with very precise prompt following. Its signature strengths are readable short text in the frame (in Latin script) and obedient composition: what you described is what you get.
Open in studio →What it can do
- ✓Precisely follows complex prompts: several objects, their placement and conditions in one request without improvising.
- ✓Text in the frame: short captions, signs and slogans in Latin script come out readable, without a "letter mess".
- ✓Editing an uploaded photo with a text instruction: only what you asked for changes, the rest is kept.
- ✓Variations and reference blending: up to 6 images in one request, the style of one + the object of another.
- ✓Realistic light, depth and faces; confidently handles multi-character scenes.
How to write a prompt
Describe the scene in a full sentence: subject → environment → lighting → style → angle. The model executes literally: what you write is what you get, so lay out the composition in words ("headline at the top, object in the center, margin on the right"). Text in the frame: short, in Latin script, in quotes. Ask for edits with a separate short instruction rather than a new prompt.
A prompt that works
Vertical 4:5 poster: a deep blue gradient, soft studio light; large in the center, wireless headphones with highlights; a large headline "PURE SOUND" at the top; a small caption "2026" at the bottom; premium minimalism, lots of air
Avoid this
“a beautiful poster with headphones and a caption in Cyrillic”
Without a layout the model will choose the composition itself, and a Cyrillic caption in the frame will come out garbled. Describe the zones in words and write the frame text in Latin script in quotes.
Examples
Banner with space for layout
Horizontal 16:9 banner: a pastel beige background, morning light; a ceramic cup of coffee with steam on the left; a large headline "NEW SEASON" on the right; lots of free space on the right for text
Editing a finished photo
(attach a photo) Make the background darker and remove the object on the left, keep the person and the light on them
Variants within the group
- Reve 2.0 Create, generation from scratch by text
- Reve 2.0 Edit, editing an uploaded photo by instruction
- Reve 2.0 Remix, variations and blending of up to 6 references
- Fast variants, faster and cheaper for drafts and iterations
Tips
- →Need a Cyrillic headline? Leave an empty zone for it and add the text as a layer in an editor.
- →Add "no extra text": the model will not paint random captions.
- →Iterate with short edits in Edit instead of rewriting the whole prompt.
FAQ
Can the results of Reve 2.0 be used commercially?
Yes, with an active Pixyn subscription. Note that some models have provider restrictions by industry, check the terms of the specific version.