CREATE

Describe the picture. Then change it.

Ask for the image in a chat and it comes back in the conversation. Need the sweater in green or the light at golden hour? Say so. The team sees the brief, the drafts and the final version in one place.

Start free trial Get the image & video playbook

What it does

Make it, then fix it.

Generating a picture is the quick part. Getting from nearly right to approved takes most of the time, so editing works the same way as asking: in a sentence.

From one sentence

Describe the picture: the product, the setting, the light. It comes back in the conversation, ready to preview and download.

Edit by saying what changes

Attach a photo and ask for the change: another colourway, a new season, different light. What you did not mention stays as it was.

Your own photos as the start

Work from the product shot you already have rather than a model's idea of your product. Most real image work is fixing a picture that is nearly right.

Your favourite writer writes the prompt

Chat models call the image tool themselves. Ask Claude Opus 5.5 for three campaign visuals and it writes three careful prompts and renders all of them.

Reviewed where the team works

Share the chat with whoever signs the picture off, so the brief, the drafts and the final version sit in one thread.

Under the workspace's rules

Admins decide which model providers the workspace may use: allow one, block one, or limit it to specific teams.

Real outputs

Straight out of the model. No retouching.

Every picture below was generated inside StickyPrompts for a fictional retailer, Northwind. Hover one to read the prompt that made it.

Prompt · Gemini 3 Pro Image Product photo for a launch banner: a sleek laptop on a warm oak desk showing a retail analytics dashboard with bright yellow and charcoal charts, soft morning window light, shallow depth of field.
Prompt · Gemini 3 Pro Image Editorial campaign photo for an autumn knitwear collection: a smiling woman in a mustard-yellow chunky sweater in a sunlit independent clothing store, warm film tones.
Prompt · Gemini 3 Pro Image Minimal social ad poster, square: bold black headline 'Every store, one screen' on a bright yellow background, a small laptop showing a clean dashboard.
Prompt · Gemini 3 Pro Image Flat vector illustration for a blog header: a friendly store manager and a laptop with rising yellow bar charts, soft paper texture, charcoal and bright yellow.

Editing, in one sentence.

Drop the photo in and say what should change. The sweater went from mustard to forest green and the light moved to golden hour; her face, her pose and the shop stayed exactly as they were. Drag the handle to compare.

Generated editorial photo of a woman in a mustard sweater in a clothing store — Edit "Change the sweater to deep forest green and make it golden hour. Keep her face, pose and the store exactly the same." Gemini 3 Pro Image

The same photo edited: the sweater is now forest green, the light is golden hour — Edit "Change the sweater to deep forest green and make it golden hour. Keep her face, pose and the store exactly the same." Gemini 3 Pro Image

ANY CHAT MODEL

Write with one model. Draw with another.

Claude Opus 5.5 has no image generation of its own. Ask it for three images anyway and it writes three careful prompts and calls StickyPrompts' image generation for each, several at once, in one reply. You keep the model you like writing with.

The model you trust with words writes the image prompt
Several images in one request, all in the conversation
Every prompt it wrote is there to read and reuse

Any model makes images. Claude Opus 5.5 has no image model. Asked for three images, it writes three prompts and StickyPrompts renders them.

IN THE CONVERSATION

Refine it by replying.

Images come back in the chat, so the next change is simply the next message. Attach the photo you have, say what should be different, and the edit lands underneath with its own preview and download. Share the thread with whoever approves it, or do the whole thing in a team chat.

Attach your own photo to edit it
Preview and download every image
Share the chat, or work in a team chat with @ai

See team chats

Image manipulation: the same photo recoloured and relit from one sentence

WHAT TEAMS USE IT FOR

Campaign and social variants from one approved hero image
Product shots moved to a new backdrop or season
Illustrations for decks that would otherwise be clip art
Blog and newsletter headers in a consistent style
Cleaning up photography that is nearly right
Concept art to brief a designer or a photographer

See every model

Straight answers

What teams ask about image generation.

Which model makes the images?

Every picture on this page was made with Gemini 3 Pro Image, straight from StickyPrompts. Which providers your workspace can use is an admin decision, set once in Model & Provider Access.

Can it keep to our brand style?

Give it a reference photo and spell out the rules - palette, framing, what never appears - and it follows them well. The practical way to make that stick is a saved prompt with your style notes and reference images attached, so nobody has to remember them. There is no automatic brand-kit lock today.

Can we use the output commercially?

That is set by the terms of the model provider you use, so check them for your case. If a provider's terms do not suit your legal team, an admin can block that provider for the whole workspace.

Where does our data go?

Workspace data is stored in the EU by default. Admins can also require EU-served inference or zero-data-retention providers only, and the settings page shows which models that leaves before you save.

Is there a guide to prompting for images?

Yes, and it is free: The AI Image & Video Playbook covers what actually changes an image and what is superstition.

Trial

A $5 balance to start. No card needed.

Bring one real brief

Try it on this week's campaign brief.

Start free with a $5 trial balance. Attach one product photo and ask for the version you actually need.

Start free trial Watch how it works Book a demo