Image generated by h3image.online with MiniMax-H3-Image from this prompt. prompt by @yu_ichi_suzuki. We never treat the author's original image as ours, or our generation as theirs.
Reference image we used
These are our own earlier MiniMax-H3-Image outputs, used as inputs for this edit. They are not the author's original reference images.
Prompt
The girl seated on the gondola bench in <Picture 1> is <Subject 1>. The Ferris wheel gondola around her in <Picture 1>, including the purple bench, the curved window frames, the ceiling lamp and the night city seen through the windows, is <Subject 2>. The drawing style of <Picture 1>, with its thin colored line art, its flat shading in soft low-contrast tones, its muted pastel palette and the way the face and eyes are drawn, is <Subject 3>. The target image shows <Subject 1> bursting out laughing in the unchanged gondola. <Subject 3> is fully preserved. <Subject 1> is partially preserved; her hair, beret, outfit and pose remain while her facial expression changes. <Subject 2> is fully preserved. Her eyes are squeezed shut in happy arcs, her mouth is wide open in a big laugh showing her upper teeth, and her cheeks are lifted, drawn in <Subject 3>. The art style and the overall atmosphere of the picture stay the same as the source.Parameters we used
- Model
- MiniMax-H3-Image
- Aspect ratio
- 3:4
- Resolution
- 880x1184
- Area
- 1K
- Task
- Image edit, 1 reference
- Generated
- 2026-11-11 11:19:46 UTC
Source
Public post on X. prompt by @yu_ichi_suzuki · 2026-11-11.
9 likes on X when captured on 2026-11-11.
Author compared MiniMax-H3-Image vs P-H3-Image on the same source illustration, locally in ComfyUI, LoRA off, 8 MP; H3-Image at 6 steps. This is MiniMax prompt 07/13 (表情（大笑い） / A big laugh), written in the official edit-guide form. The P-H3 wording from the same post is not included. Comparison video/images are in the parent post. Link only.
