Mira image prompting
The image pipeline works like the video one: your facts are kept, a planner and four specialists
fill the light, optics, composition and grade you did not state, a critic projects the result into
the model's format. So the job is to be exact about the subject and clear about the one or two
craft decisions you care about.
Step 1. Choose the model
Name the ONE thing that decides the image:
| The image needs |
Model |
| text in frame, any language, Cyrillic |
gpt-image-2.5-flare (or nano-banana-pro) |
| a polished campaign visual or a faithful edit of the user's photo, inspected closely |
gpt-image-2.5-sunburst (premium) |
| the same product or person across several images |
seedream-4 |
| print-grade detail, a large-format hero |
seedream-4k |
| an exact tall ratio (9:16, 4:5) that must come back exactly |
gpt-image-2.5-flare, nano-banana-pro or seedream-4 |
| fast drafts and mood exploration, no text |
flux |
| everything else: a detailed commercial brief, packshots, materials, skin |
gpt-image-2 |
Costs and quirks per model: references/models.md.
Step 2. Build the prompt
Read references/craft.md, generated from the registry, once per session.
One paragraph of prose, 90 to 160 words, in this order: subject with material detail; composition
and framing for the exact ratio; background and depth; the lighting setup with a source and a
direction; palette; optics and shot size; surface and texture; grade; the empty part of the frame
left for later. Choose one lighting scheme, one optics entry, one grade and at most two materials.
Step 3. Pick from the banks, write in words
The banks are for choosing. Open the one you need, pick an entry, and write its description into
the prompt in plain words. A key such as three_point or macro_100 in a prompt is a bug.
| Decision |
Read |
| light |
references/light.md |
| lens and shot size |
references/optics.md |
| how the frame is arranged |
references/composition.md |
| colour treatment |
references/grade.md |
| surfaces and materials |
references/materiality.md |
Step 4. Check
- Every adjective is checkable against the finished image. Delete the ones that are not.
- No tag soup, no brand shorthand, no praise words. See the banned list in
references/craft.md.
- The product's label and the location's signage are the user's own and stay exact; third-party
brands are described generically.
- Text appears in double quotes, exactly, or not at all.
Output
Model: <id>
Ratio: <w:h>
Prompt: <one paragraph>
1---2name: mira-image-prompting3description: How to write stills for Mira AI's image models: GPT Image 2, Nano Banana Pro, Seedream 4 and 4K, Flux. Which model holds text, which keeps a product identical across a series, how the craft banks for light, optics, composition, grade and materials are used, and the words that make an image worse. Use it before any generate_image call on Mira, or when the user says "image prompt", "промпт для картинки", "нарисуй", "сгенерируй картинку", poster, packshot, banner, cover. NOT for video (mira-video-prompting).4license: MIT5---67# Mira image prompting89The image pipeline works like the video one: your facts are kept, a planner and four specialists10fill the light, optics, composition and grade you did not state, a critic projects the result into11the model's format. So the job is to be exact about the subject and clear about the one or two12craft decisions you care about.1314## Step 1. Choose the model1516Name the ONE thing that decides the image:1718| The image needs | Model |19|---|---|20| text in frame, any language, Cyrillic | `gpt-image-2.5-flare` (or `nano-banana-pro`) |21| a polished campaign visual or a faithful edit of the user's photo, inspected closely | `gpt-image-2.5-sunburst` (premium) |22| the same product or person across several images | `seedream-4` |23| print-grade detail, a large-format hero | `seedream-4k` |24| an exact tall ratio (9:16, 4:5) that must come back exactly | `gpt-image-2.5-flare`, `nano-banana-pro` or `seedream-4` |25| fast drafts and mood exploration, no text | `flux` |26| everything else: a detailed commercial brief, packshots, materials, skin | `gpt-image-2` |2728Costs and quirks per model: [references/models.md](references/models.md).2930## Step 2. Build the prompt3132Read [references/craft.md](references/craft.md), generated from the registry, once per session.33One paragraph of prose, 90 to 160 words, in this order: subject with material detail; composition34and framing for the exact ratio; background and depth; the lighting setup with a source and a35direction; palette; optics and shot size; surface and texture; grade; the empty part of the frame36left for later. Choose one lighting scheme, one optics entry, one grade and at most two materials.3738## Step 3. Pick from the banks, write in words3940The banks are for choosing. Open the one you need, pick an entry, and write its description into41the prompt in plain words. A key such as `three_point` or `macro_100` in a prompt is a bug.4243| Decision | Read |44|---|---|45| light | [references/light.md](references/light.md) |46| lens and shot size | [references/optics.md](references/optics.md) |47| how the frame is arranged | [references/composition.md](references/composition.md) |48| colour treatment | [references/grade.md](references/grade.md) |49| surfaces and materials | [references/materiality.md](references/materiality.md) |5051## Step 4. Check5253- Every adjective is checkable against the finished image. Delete the ones that are not.54- No tag soup, no brand shorthand, no praise words. See the banned list in55 [references/craft.md](references/craft.md).56- The product's label and the location's signage are the user's own and stay exact; third-party57 brands are described generically.58- Text appears in double quotes, exactly, or not at all.5960## Output6162```63Model: <id>64Ratio: <w:h>65Prompt: <one paragraph>66```
Global-AI-Group-Inc