Routes each creative deliverable to the right AI image or video model by its failure mode - typography, vector, photoreal with brand hex, campaign consistency, dialogue, human motion - and compiles one creative spec into each model family's prompt dialect. Use when someone asks "which AI model should I use for product images", "why does my generated text come out as gibberish", "which AI image model fits each asset type in our campaign", "how do I turn one campaign brief into prompts for different models", or when a launch needs several asset types generated and nobody has picked tools. Do NOT use for writing the individual prompt - use photoreal-prompt-craft instead; for text-only LLM prompting, use prompt-engineer.
---
name: image-model-router
description: Routes each creative deliverable to the right AI image or video model by its failure mode - typography, vector, photoreal with brand hex, campaign consistency, dialogue, human motion - and compiles one creative spec into each model family's prompt dialect. Use when someone asks "which AI model should I use for product images", "why does my generated text come out as gibberish", "which AI image model fits each asset type in our campaign", "how do I turn one campaign brief into prompts for different models", or when a launch needs several asset types generated and nobody has picked tools. Do NOT use for writing the individual prompt - use photoreal-prompt-craft instead; for text-only LLM prompting, use prompt-engineer.
---
# image-model-router
One-model monoculture forces every deliverable through one failure mode - the router exists because models fail differently: forcing an aesthetic-exploration model to render an infographic, or a chat UI to do layout-exact work, produces the gibberish text and drifting brand color that out the whole launch as generated. Leaderboard rank is task-conditional - a top-3 general model can be the worst choice for a specific deliverable. You are the dispatcher. Nothing renders until it has a lane.
## Operating procedure
Run the steps in order. The manifest must exist before any routing decision, routing before dialect compilation, dialect before handoff - each step consumes the previous step's table.
### Step 1: Elicit the client charter and the deliverable manifest
Gather exactly these inputs before routing anything. If a number is a guess, label it a guess and move on.
1. **Charter**: the client's brand charter block. Paste the charter block if one exists. If the client has a charter but the foundations pack is not installed, elicit the needed values directly and label them provisional. If no charter exists at all, halt and route to premium-design-foundations - never substitute worked-example values.
2. **Assets**: every asset the campaign needs, as artifact types (packshot, macro detail, lifestyle, gallery frame, social still, logo or vector, clip, trailer) with target dimensions and aspect ratios.
3. **Text**: which assets carry in-image text, with the word count per text element.
4. **Color**: brand color requirements as hex.
5. **Indemnification**: whether legal indemnification is required.
6. **Deadline and budget posture**: draft-tier exploration allowed, or premium-only.
7. **Set membership**: which assets must match each other.
8. **UI beats**: which video beats show product UI - those never enter a video lane (Step 3).