Start Free

Review & Feedback

Browse your generated images, leave targeted feedback, and let the AI automatically improve your style guide using vision.

Browsing Images

The Review page displays all generated images grouped by batch (generation runs separated by time). For each batch you can toggle between:

  • •Original — Images with their original white background
  • •Transparent — Background-removed versions shown on a checkered pattern

Hover over any image to zoom in for a closer look.

Per-Image Feedback

Each image has a text field where you can leave specific feedback about what needs improvement. Be as specific as possible:

✓ "Bun is too glossy and shiny"

✗ "Looks weird"

✓ "Cheese needs more visible melt"

✗ "Fix the cheese"

✓ "Angle should be more overhead"

✗ "Wrong angle"

✓ "Colors are oversaturated, too warm"

✗ "Colors are off"

You can also redo any single image (regenerate with the current style guide) or remove it entirely (archives the image).

Bulk Feedback

When multiple images share the same issue, use bulk feedback instead of writing the same note on each one:

  1. 1.Select multiple images using the checkboxes
  2. 2.Write a joint feedback note (e.g., "All these need more vibrant colors and less shadow")
  3. 3.The AI expands the joint note into specific per-image feedback tailored to each item

The Vision Feedback Loop

This is Asset Forge's most powerful feature. When you click Revise with AI, the system:

1

Loads Your Images

The actual PNG files are loaded and encoded to base64 — the AI will see the real images, not just descriptions.
2

Sends Images + Feedback to AI

Each image is paired with its text feedback and sent to a vision model. The AI can see exactly what's wrong — a glossy bun, an unnatural shadow, oversaturated colors.
3

AI Revises the Style Guide

Based on what it sees and your feedback, the AI makes targeted changes to the style guide. The old version is saved to history.
4

Images Are Regenerated

All images that had feedback are regenerated with the improved style guide. The old versions are archived.
The vision feedback loop uses a stronger model than basic prompt writing. Revisions don't cost credits — they're included to help you improve your results.

Overall Notes

In addition to per-image feedback, you can write overall session notes that apply to all images. These are included when the AI revises the style guide, providing high-level direction like "all drinks should be photographed from the side" or "reduce garnish across the board."