Fashion AI SchoolLab

Google Gemini · Guide

Google Gemini for visuals and research

Gemini is strongest when your work touches images and your own files at the same time: editing a product photo, keeping one subject consistent across scenes, or researching across documents you already keep in Google. Use it for image editing and grounded research rather than long-form writing.

What it is good at

  • Image editing with plain language — swap a background, change light, place a product in a scene.
  • Understands what is in your uploaded photo and edits it rather than generating something similar.
  • Deep research runs across many sources and returns a report with links.
  • Sits next to your documents, so research and drafting share the same material.

Where it falls short

  • Written voice needs more correction than Claude.
  • Image results vary between attempts — expect to run several and pick.
  • Fine typography and small logo detail still break.

How to use it for creative work

  1. 01

    Start from a real photo

    Upload the actual product or garment. Editing a real image beats generating a lookalike every time.

  2. 02

    Describe the change, not the picture

    Say what should be different — 'same shoe, morning window light, concrete surface' — instead of re-describing the whole shot.

  3. 03

    Lock the subject

    Repeat the identity sentence in every request: 'the same [item], unchanged shape, colour and stitching'. Consistency comes from repetition.

  4. 04

    Run research grounded in sources

    Ask for a report with links and dates, and tell it to mark anything it could not verify. Unsourced trend claims are the main risk.

  5. 05

    Export before you iterate

    Save each good frame immediately; long editing chains drift away from the original product.

Ready-to-use prompts

Clean product relight
Using the attached photo: keep the product identical in shape, colour and detail. Replace the background with a plain warm stone surface and light it with soft directional morning light from the left. Realistic contact shadow. No props.
Same subject, new scene
Same subject as the attached image, unchanged. Place it in [scene]. Match the perspective and the height of the camera to the original. Keep materials and colours exactly as they are.
Scene set of five
Produce five variations of the attached image in one consistent visual world: [describe world]. Vary only the angle and the light. The subject must be identical in all five.
Grounded research report
Research [topic] across the last six months. Return a report with: eight findings, each with a source link and date; three implications for a small independent brand; and a clearly marked list of claims you could not verify.
Document comparison
Compare the attached documents. Return: where they agree, where they conflict, and which conflicts actually change a decision. Be specific, quote the lines.

Questions people ask

Is Gemini good enough to replace a photoshoot?
For flat-lay variations, backgrounds, seasonal reshoots and social crops, often yes. For hero campaign imagery where texture and fit must be exact, it still supports a shoot rather than replacing it.
Why does my product change slightly each time?
Because each run is a fresh generation. Repeat an identity sentence, edit from the same source image, and keep the number of changes per request to one.

Get free workshops, tools and prompts

Occasional updates from Fashion AI School: free live workshops, new AI tools and new prompts. No spam, unsubscribe any time.