# Bria AI Reimagine Model

## What is Bria AI Reimagine?
Bria AI Reimagine is an advanced text-to-image generation model in the Bria AI suite that lets you preserve a source layout while applying new styles and visual treatments. Unlike Base, Fast, or HD variants that focus on speed or fidelity, Reimagine uses a structure reference image (via ControlNet) to guide composition, depth, and edges so you get consistency in scene structure with full creative flexibility on texture, color, and lighting.

## Key Features
- **Structure Reference Control**  
  – Use `structure_image_url` or `structure_image_file` to anchor your composition  
  – `structure_ref_influence` slider (0–1) controls adherence to the reference  
- **ControlNet Guidance Methods**  
  – Edge detection, depth maps, recoloring, color grids for fine-grained control  
- **Speed vs. Quality Modes**  
  – Fast Mode (boolean): prioritize throughput for quick iterations  
  – HD Mode: maximize detail with higher-resolution output  
- **Reproducibility & Refinement**  
  – `seed` parameter ensures identical results across runs  
  – `steps_num` (4–50) sets diffusion iterations for more or less detail  
- **Optional Enhancements & Safety**  
  – `enhance_image`: boost clarity and sharpness  
  – `prompt_content_moderation` & `content_moderation` toggles enforce brand and safety policies  
  – `ip_signal` detection for potential copyright issues  

## Best Use Cases
- **Creative Storyboarding**: Maintain camera angles and scene blocks while iterating on color and style.  
- **Product Mockups & UI Layouts**: Preserve layout grids or wireframes and restyle for different themes.  
- **Illustration & Concept Art**: Generate consistent character or environment sketches with unique artistic treatments.  
- **Marketing & Social Media Assets**: Produce templated visuals that align across campaigns with varied moods.  

## Prompt Tips and Output Quality
- **Be Descriptive**: Detailed prompts yield richer textures and more accurate scene elements.  
- **High-Resolution References**: Supply clear structure images to avoid artifacts in complex compositions.  
- **Adjust Influence**:  
  - Lower `structure_ref_influence` (~0.3–0.5) for looser interpretation  
  - Higher values (>0.8) for strict adherence to the reference  
- **Tweak Refinement**: Increase `steps_num` (20–30) for ultra-fine details; lower it (8–12) for rough drafts.  
- **Seed Consistency**: Use the same `seed` to reproduce previous outputs exactly, perfect for A/B testing.  
- **Moderation Settings**: Toggle off moderation flags if you require unfiltered creative exploration.  

## FAQs

**Q: How do I maintain the original layout of my reference image?**  
A: Upload a clear base image via `structure_image_file` or URL and set `structure_ref_influence` high (0.75–1.0).

**Q: Can I speed up rendering without sacrificing all detail?**  
A: Enable `fast` mode for quicker drafts, then rerun in HD mode or bump up `steps_num` for final versions.

**Q: What’s the default seed and steps?**  
A: By default, `seed` is 42 and `steps_num` is 12. Adjust both for reproducibility and quality control.

**Q: How can I enforce brand-safe outputs?**  
A: Turn on `prompt_content_moderation` and `content_moderation` to filter generated images against unsafe or IP-sensitive content.

**Q: Is custom color control supported?**  
A: Yes—combine ControlNet color grid guidance with descriptive prompts for consistent color theming.