Skip to content

Global-AI-Group-Inc/content-generation-skills / skills/mira-image-prompting

mira-image-prompting

Verified

How to write stills for Mira AI's image models: GPT Image 2, 2.5 and 2.6, Nano Banana Pro and 2.1, Seedream 4, 4K and 5 Pro, Muse Image, Ideogram 4.5, 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).

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 needsModel
text in frame, any language, Cyrillicgpt-image-2.6-flare (or nano-banana-2.1, nano-banana-pro)
a polished campaign visual or a faithful edit of the user's photo, inspected closelygpt-image-2.6-sunburst (premium)
the same product or person across several imagesseedream-4
premium photoreal at true 2K: skin, fabric, glass, metal that must look photographedseedream-5-pro (4 credits)
an everyday photoreal still, or a step-by-step edit of one photo that must leave everything else untouched; a sticker or cut-out on a transparent backgroundseedream-5.5 (2 credits; no 4:5 or 5:4)
a real, specific place, landmark or product that must be recognisable; cheap draftsmuse-image (1 credit)
a poster, cover, label or layout where the lettering IS the designideogram-4.5 (6 credits)
a panoramic banner, header or strip (21:9, 4:1, 8:1) or a tall scroll (1:4, 1:8)nano-banana-2.1 (4 credits, the only model with these ratios)
print-grade detail, a large-format heroseedream-4k
an exact tall ratio (9:16, 4:5) that must come back exactlygpt-image-2.6-flare, nano-banana-2.1, nano-banana-pro, seedream-5-pro, ideogram-4.5 or seedream-4
fast drafts and mood exploration, no textflux
everything else: a detailed commercial brief, packshots, materials, skingpt-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.

DecisionRead
lightreferences/light.md
lens and shot sizereferences/optics.md
how the frame is arrangedreferences/composition.md
colour treatmentreferences/grade.md
surfaces and materialsreferences/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>
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall