Training Your AI with Feedback

Every time the AI generates or edits a thumbnail for you, you can give it feedback. Over time, this trains the AI to understand your preferences and produce better results.

How Feedback Works

The feedback system has two levels:

Quick Feedback (Thumbs Up / Down)

After any AI generation or edit, you'll see 👍 thumbs up and 👎 thumbs down buttons on the result. A quick tap records whether you liked or disliked the result.

Detailed Feedback

When you give a thumbs down, a sheet slides up where you can provide more specific information:

• Categories: Select what went wrong — face accuracy, colors, text placement, lighting, style, composition, or other
• Text feedback: Describe exactly what you'd like changed in your own words

This detailed feedback is especially valuable when the AI gets something specific wrong, like generating the wrong face (which happens roughly 1 in 20 times) or using the wrong colors.

Giving Effective Feedback

What the AI learns from thumbs up

A thumbs up tells the system "more like this." It reinforces:

• The overall style and mood
• Color choices
• Composition and layout
• Text placement and typography

What the AI learns from thumbs down + details

Detailed negative feedback teaches the AI to avoid specific mistakes:

• "The face doesn't look like me" — helps improve face accuracy
• "The background is too busy" — teaches background preferences
• "Text is too small and hard to read" — improves typography decisions
• "Colors are too dark/saturated" — refines color choices

Tips for better feedback

• Be specific — "the face is wrong" is more useful than just a thumbs down
• Select the right category — this helps the system route feedback correctly
• Give feedback on every generation when possible — even thumbs up matters
• For face issues, mention it explicitly — face accuracy is tracked separately

How the Learning System Works

Your feedback feeds into a learning pipeline that:

1. Records each positive and negative signal
2. Analyzes patterns across your feedback history
3. Injects learnings into future prompts — so the AI remembers what you like
4. Works per-organization — your whole team benefits from the feedback

Combined with Design Rules (always-on constraints) and Styles (visual templates), the feedback system gives you a third layer of personalization that improves automatically over time.

Troubleshooting

If you see "Failed to save feedback," try:

1. Refresh the page and try again
2. Make sure you've selected at least one category when giving detailed feedback
3. Check that your feedback text isn't empty if you've opened the details panel

The system validates feedback categories against a fixed list, so if you encounter errors, it's typically a temporary sync issue that resolves on refresh.