The starting point: you describe what you want, the agent plans, generates and files the results in a project.
Steps
Describe what you needDrop your packshots and write the request in your own language: use of the visual, mood, format, product dimensions. The agent improves simple prompts on its own.
One chat per jobSame chat for the same process on similar products; new chat for another product or another style.
Use the shortcuts/ to run a skill, add to insert an Element.
Find the resultImages land in a project, which you can then send for review and download.
FAQ
What should go in the first message?
The product, with a good-quality packshot.
The use: product page, social media, print.
The expected mood and format.
The product dimensions if size matters in the scene.
Should I open a new chat for each image?
Same chat for the same process on similar products (10 sofas, same style).
New chat for a different job: another product, another style. A chat that runs too long ends up mixing instructions.
How many options should I generate?
Two options are enough in most cases. Each option uses credits: the cost is shown before each run.
Project, item, output: what is the agent talking about?
A project groups a job, an item is a product in that project, an output is an image generated for that item. “Look at all accepted outputs in this project” is understood without ambiguity.
Can I write my prompts in my own language?
Yes, the agent understands all languages, with no impact on quality.
A skill is a saved set of instructions, shared with your whole organization. It saves you from rewriting the same instructions and keeps results consistent from one person to the next.
Steps
Get one product rightWork on a single product until the result is right.
Save itWrite “save this process as a skill”. You can also start from scratch or from an existing Presti skill.
Test and adjustRun it on 3 to 5 other products; for each gap: “update the skill with…”.
Run itType / followed by the skill name.
FAQ
How do I adapt a Presti skill?
Write: “create a new skill based on /FabricSwap with these rules…”. The original skill stays intact.
How do I organize skills across the team?
Skills are shared with the whole organization: set a naming convention from the start (brand, category, use).
Start with your priority category, then duplicate and adapt.
How do I move a skill from one workspace to another?
Download it, switch workspace, then drop it into the agent.
Does a skill work for video?
Yes: a video format that works can be replayed across the whole range in the same way.
Review mode is for approving a project’s visuals, alone or as a team, and sending your corrections to the agent.
Steps
Open the reviewFrom the project, open Review mode to go through the outputs one by one.
Accept or rejectEach decision updates the output’s status; then filter on accepted outputs to deliver.
CommentOne precise comment per visual: “reduce the sofa size by 20%”.
Have the agent fix itIn the chat: “apply the comments in this project”. The agent handles all comments at once.
FAQ
How do I fix 50 visuals with the same instruction?
Add the same comment to each visual concerned (“reduce size”), then in the agent: “find the visuals commented ‘reduce size’ and reduce the product by 20%”.
How do I get a colleague to approve?
Project page, Request a review, pick the person: they receive an email with the project link.
Rerun from a comment or edit the image?
Rerunning from comments starts from a clean generation. It is often better than several successive edits, which degrade the image.
The AI Assistant is for small fixes on an almost perfect image. It starts from a blank context, without the agent’s chat history.
Steps
Open the imageClick the image in the project, then open the AI Assistant.
One fix at a timeFor example: “keep the scene identical, enlarge the sofa by 20%”.
Add a reference if neededA reference image helps keep the product’s exact color or shape.
Limit editsOne edit per image, two at most: beyond that, quality degrades.
FAQ
What’s the difference with the agent?
The agent keeps the full chat history: ideal for creating, less so for fine retouching. The AI Assistant starts from a blank context and doesn’t mix your request with previous instructions, so the result is more faithful.
The result drifts after two tries: what should I do?
Go back to the original image and rephrase, rather than retouching the result.
My product is too big or too small
“Keep the scene identical, enlarge the product by 20%”, giving its real dimensions if possible.
Restore Colors reapplies the colors of a reference image to an image that has turned pink. The tool uses no credits.
Steps
Open the tinted imageIn the project, open the image, then the arrow at the top and Restore Colors.
Choose the referenceSelect the image with the right colors, usually the original version.
SaveSave the result in the project.
Check the productThe correction applies to the whole image: check the product’s color.
FAQ
Why does my image turn pink?
It’s pink drift, a known model limitation: a pink tint appears over successive edits.
Does Restore Colors use credits?
No, the correction is free.
The product color changed after correction
The correction applies to the whole image and can slightly alter the product. If the product is no longer accurate, start again from a clean generation.
Organize a multi-product project with a structured file
On a project with many products, a structured file gives the agent a clear reference: each product, its image, its information.
Steps
Have the file createdDrop your images and ask the agent to organize them in an Excel or CSV file: product name and packshot link. The agent turns each image into a URL.
Complete itAdd dimensions or other information, and ask the agent to add them to the file.
Generate from the fileAsk for lifestyles based on the file’s content: since the information is already structured, the agent doesn’t mix up products.
FAQ
Can I create Elements in bulk the same way?
Yes: drop a table (name, description, image links) into the agent to create all the Elements at once.