A design is a creative_studio record: the canvas, its frames (the artboards, one per size) and its brand kit. The canvas lives in the editor in the person's browser. The AI changes it with the cs_* tools, and only while the design is open on their screen. Each tool waits for the editor to confirm what it did. The AI can save a design. It can't export one: the person presses Export.
Ask first
- Where will it be posted? The platform and the format (post, portrait, story). This sets the size.
- The words: the headline, the line under it, the call to action, any price or date. Use only what they give. Never invent an offer.
- The picture: one of theirs (a file in Files, or one they upload now), or a generated one. Generating costs money (see the Stop in step 4).
- The brand: is there a brand design (see Brand kit for designs)? If not, their colours (hex) and fonts.
- Where does it go after? Downloaded, kept in Files, or attached to a draft post (which platform).
Steps
1. Creative Studio library: the design
Use the Creative Studio endpoints through call_appengine, not create_record. The create endpoint makes the unique name and the frame.
- With a brand design: find it with
POST /creative-studio/designs/list{ "search": "brand" }. Copy it withPOST /creative-studio/designs/<brand sk>/duplicate, then name the copy withPOST /creative-studio/designs/<copy sk>/rename{ "title": "<name>" }. The copy keeps the brand design's size. If the size is wrong, change it in step 3. - Without one:
POST /creative-studio/designs/create{ "title": "<name>", "size": { "width", "height", "name", "platform" } }. Each side must be 16 to 10000 pixels, or the answer is 400.
The sizes the New design drawer offers:
| Where | name | Size (px) |
|---|---|---|
| Instagram post | Instagram Post | 1080 × 1080 |
| Instagram portrait | Instagram Portrait | 1080 × 1350 |
| Instagram story or reel, TikTok, YouTube Short | Instagram Story | 1080 × 1920 |
| Facebook post | Facebook Post | 1200 × 630 |
| X post | X / Twitter Post | 1200 × 675 |
| LinkedIn post | LinkedIn Post | 1200 × 627 |
| Pinterest pin | Pinterest Pin | 1000 × 1500 |
| YouTube thumbnail | YouTube Thumbnail | 1280 × 720 |
| Facebook cover | Facebook Cover | 820 × 312 |
| Web banner | Web Banner | 728 × 90 |
| Flyer (print, 300 dpi) | US Letter / A4 | 2550 × 3300 / 2480 × 3508 |
Hands on: the design's sk.
Check: the design's card is in the library (/creative-studio). Search for its name.
2. Creative Studio editor: open it
navigate { route: "/creative-studio", params: { uid: "<design sk>" } }. Creative Studio has no entry in find_app, so navigate to it directly.
Do the design work in the editor's Design AI bar. It sends the canvas with each request: the frames, every object with its id, the brand kit and what is selected. A request from another chat doesn't carry the canvas, so the AI can't see what is on it.
Check: the editor shows the design and its frame.
3. The brand kit and the size
- A copy of the brand design already has the kit (
context.canvas.designSystem). - Otherwise
cs_apply_design_system{ name, colors: { primary, secondary, … }, headlineFont, bodyFont }. Fonts are Google Fonts names. - Wrong size:
cs_resize_frame{ width, height, scaleContent: true }.
Check: the brand kit panel (the palette button, top right) shows the colours and fonts.
4. The picture
| Picture | How |
|---|---|
| Theirs, already hosted (Files, an upload) | cs_add_to_canvas { objects: [{ type: "image", url: "<https url>", fit: "cover" }] }. With no box, cover fills the frame. |
| Generated | cs_generate_image { prompt, placement: "background" }. The image covers the frame, behind everything. |
| Theirs, changed by AI | The person selects it, then cs_generate_image { prompt, referenceImage: "selection", replaceSelection: true }. |
| Theirs, cut out | The person selects it, then cs_remove_background. The cut-out replaces it in place. This uses Appmint's own background remover and is not billed as AI image generation. |
The AI can't select objects. Only the person can, by clicking on the canvas. A picture pasted from their computer, not hosted, can't be cut out or used as a reference. Ask them to upload it again.
Stop: every generated image is billed to the business: per image on Appmint's AI (see Service rates), or by their own provider if they use their own AI key. Say what will be generated, and that it costs, then wait for yes. Make one image per yes. Don't retry or make variations without asking.
A generated image is saved in Files under ai-generated/.
5. The words
Add text with cs_add_to_canvas. Each line is a textbox with text, fontSize, fontFamily, fontWeight, fill, textAlign and width. Coordinates count from the frame's top-left corner.
- Headline about frame width ÷ 10, body about frame width ÷ 30. Centred text:
left: 60,width: frame width − 120,textAlign: "center". - Light text on a dark picture and dark text on a light one.
- Add the text after the background picture, so it sits on top.
To change something later, use cs_update_objects with the ids from context.canvas.objects, or "selected" for what the person selected.
Hands on: the ids of what was added.
Check: every tool answered confirmed: true. If one didn't, tell the person what did not happen.
6. Save
cs_save { title? } saves the canvas, the frames, the brand kit and a preview for the library. The editor also saves by itself.
Check: the library card shows the new preview. POST /creative-studio/designs/list { "search": "<name>" } shows hasContent: true and a thumbnail.
7. Export: the person does it
No tool exports. Ask the person to press Export (the download button in the bottom bar):
| Choice | Options |
|---|---|
| What | One frame, all frames, or the selection |
| Format | PNG, JPEG (with a quality), or SVG (one frame only) |
| Size | 1×, 2× or 3× the frame. 1× is the platform's size. |
| Save to | This computer, or Files (folder creative-studio/exports unless they change it) |
Then find the file with call_appengine POST /repository/file/flatlist { "prefix": "creative-studio/exports/" }. Keep the trailing slash. Each file has its path and signedUrl. The file is named after the design, and with several frames, after the design and each frame.
A picture from a website that blocks other sites from reading its images stops the export ("An image on the canvas comes from a host that blocks exporting"). Upload the picture to Files, put that copy on the canvas, and export again.
Hands on: the file's url.
8. Marketing & Ads: the draft post (if asked)
marketing_draft_social_post { platform, text, imageUrl: "<file url>", campaignId? }. It saves a social post as a message with deliveryType = the platform, status: "draft" and the picture in files. No account is chosen yet. Platforms: facebook, instagram, linkedin, twitter, tiktok, pinterest. The text is plain, with no Markdown, within the platform's limit.
Stop: never publish or schedule it. The person opens it in Social Media → Posts (/social-manager?tab=posts), picks the account under Post to account and presses Send or Schedule (see Social posts and the social inbox).
Check: the draft is in Social Media → Posts with the picture on it.
When it's done
Tell the person:
- the design's name and size, and that it is in the Creative Studio library;
- where the picture came from, and whether an image was generated (and billed);
- the exported file in Files, if they exported it;
- the draft post and its platform, if one was made.
Open the design in the editor.
What it can't do yet
- No export by the AI. The AI can save, not export. The library preview is at most 640 pixels: it is not a file to post.
- No video. A design is a still picture. The story, reel and TikTok sizes make a 1080 × 1920 image. The engine has a video endpoint (
POST /ai/generate/video), but no screen and no AI tool uses it. Don't call it. - No selecting. Cutting out or reworking a picture works on what the person selected. The AI can't select anything itself.
- Outside Creative Studio,
generate_imagewith no reference image doesn't save the result. It returns the provider's link, not a file in Files. In Creative Studio, usecs_generate_image, which saves toai-generated/. - Deleting a design leaves its preview file in Files under
creative-studio/thumbnails/.
Don'ts
- Don't create or change designs with
create_record,repo_createorupdate_record. Use the Creative Studio endpoints andcs_save. Saving ignores every field except the canvas, frames, viewport, brand kit, link, platform, format and tags. A design'sstatuscan't be set. - Don't write canvas JSON by hand through
designs/<sk>/save. Lay the design out in the editor. - Don't generate an image without a yes, and don't make several to choose from unless asked. Each one is billed.
- Don't put prices, dates or discounts on the graphic that the person didn't give.
- Don't use pictures from websites the business doesn't own. Use their files and uploads.
- Don't say something is on the canvas unless the tool answered
confirmed: true.