Creatives

A creative is one ad image: a picture, with your headline, logo and button laid over it by a template, rendered in every size the campaign’s platform takes. The AI can draw the picture, or you upload your own, or you use no picture at all. The words are never drawn by the AI: a template adds them on your computer, so they’re always spelled right.

This page covers the Creatives tab: the sizes, making and editing a creative, suggested ideas, the logo, what drawing costs and how the images are rendered. To plan the campaign first, see Ads.

Find it

Open a campaign in Ads. The tabs under its name are Plan and ads, Creatives and Results. Creatives lists the campaign’s creatives, newest first, with the buttons:

• New creative opens the creative dialog.
• Suggest ideas asks the AI for three ideas from the campaign’s ads. See Suggested ideas. It’s disabled until the campaign has ad groups.
• Upload logo (or Replace logo and Remove) on the right. See The logo.

Sizes for each platform

The Sizes card lists the sizes Startica renders for the campaign’s platform and the file limit each image is kept under. The sizes were last checked against the platforms’ help pages on 2026-10-10.

Platform	Sizes	File limit
Google Ads	1.91:1 at 1200×628 (Landscape: Display, Demand Gen, Performance Max), 1:1 at 1200×1200 (Square), 4:5 at 960×1200 (Portrait: Demand Gen), 9:16 at 1080×1920 (Vertical: YouTube Shorts, Demand Gen)	5 MB
Meta	4:5 at 1440×1800 (Feed, recommended), 1:1 at 1440×1440 (Feed), 9:16 at 1440×2560 (Stories and Reels), 1.91:1 at 1200×628 (Links and right column)	30 MB
Reddit	4:5 at 1080×1350 (Feed, the tallest Reddit shows), 1:1 at 1080×1080 (Feed)	3 MB

Labelling AI images

The card also says what each platform asks about AI-made images:

Platform	What it asks
Google Ads	Disclosure of AI-made images in election ads, and since July 2026 for AI assets shown in the EU, India and New York: check the AI-content setting when you upload there.
Meta	Disclosure only in political and social-issue ads; Meta labels AI media it detects itself (since June 2026).
Reddit	It follows its political-ads rules for disclosure; check its ad policy before using AI-made people or places in sensitive topics.

These notes aren’t legal advice. Each platform’s own policy and review decide.

Make a creative

Click New creative. The dialog has the fields on the left and a live square preview on the right.

Field	What to enter
Picture	Draw it with AI, Upload my own, or No picture, just the words. See The picture.
What the picture shows	With Draw it with AI: the scene, up to 1,000 characters. For example “A field researcher’s notebook and coffee on a sunlit wooden table, soft morning light”. Describe the scene, not the words.
Image	With Upload my own: a PNG, JPEG or WebP file, up to 20 MB.
Headline	Required, up to 80 characters.
Supporting line (optional)	Up to 140 characters.
Button (optional)	Up to 24 characters, for example “Try it free”.
Price or offer (optional)	Up to 24 characters, for example “$9/mo”. Shown as a tag in the top corner.
Template	Overlay, Panel or Frame. See Templates.
Background, Words, Button	The three colors. They start as #111827 (near black), #ffffff and #f59e0b (amber).
Add the logo	Ticked by default. Adds the product’s logo, if you’ve uploaded one.
For ad (optional)	The whole campaign, or one of its ads (shown as “group: ad n”). Only a label to help you match images to ads.

Click Draw and render (with Draw it with AI) or Render. The creative appears in the list with its status, and the page updates by itself as it finishes.

The preview shows the template at 1080×1080 as you type, with your upload or the current picture. An AI picture appears only once it’s drawn.

The picture

Choice	What happens	Cost
Draw it with AI	OpenAI’s image model draws the picture from what you wrote, then the sizes are rendered. Takes about a minute.	One picture, about $0.05 at the default quality. See What drawing costs.
Upload my own	Your file is used as it is, then the sizes are rendered.	Nothing
No picture, just the words	The template draws the words on the background color.	Nothing

Draw it with AI needs an OpenAI API key, even if Claude is your provider; without one it’s disabled with “needs an OpenAI key in Settings”. See The image task.

What you wrote can’t name a competitor from the product’s Intel: that’s refused with “Leave name out of the picture: ads can’t show other brands”.

Each size crops the picture from the middle to fill its space, so a wide size loses the top and bottom of a square picture and a tall size loses the sides. AI pictures are drawn square and large for that reason. For uploads, square or larger works best, with the subject in the middle.

Templates

Template	Layout
Overlay	The picture fills the ad; the words sit on a shaded edge: along the bottom in square and tall sizes, down the left side in wide ones.
Panel	The picture on one side, the words on a solid panel of the background color: the picture on top in square and tall sizes, on the right in wide ones.
Frame	A solid background, the words above and the picture framed below (side by side in wide sizes), with rounded corners and a shadow.

The words are sized from each ad’s shorter side, so they look the same in every size. If the headline, supporting line and button don’t fit their space, the template shrinks them step by step, down to 40% of their starting size. The button keeps the background color for its text and the button color behind it.

If the words had to shrink below 70% of their size, or still don’t fit at 40%, the creative says so above its downloads, with the sizes affected: “The words had to shrink a lot to fit at 1.91:1, so they may be hard to read.” or “The words don’t fit at 1.91:1: they’re cut off.” A size whose words are cut off shows a warning sign on its download button. Shorten the headline or supporting line, or try another template; wide sizes (1.91:1) have the least room.

Suggested ideas

Suggest ideas asks the AI for three creatives, each testing a different angle: the problem, the outcome, the product in use, or a scene from the customer’s day. It reads the product’s details, the campaign’s goal and plan, and its ads, and writes for each idea:

• what the picture shows, for the image model;
• a headline of up to about 40 characters, a supporting line, and a button;
• a template, and three colors (words with strong contrast on the background, and a button color that stands out from both);
• which ad it’s for, or the whole campaign.

Ideas appear with the status Idea and “Not drawn yet”. No picture is drawn and nothing is spent on images until you click Draw it on one (the button shows the estimate). Edit or upload opens the dialog to change the idea or use your own picture instead.

Suggesting ideas uses the draft task on your usual provider, and costs a few cents. An idea whose picture names a competitor is dropped. While it runs, the button reads Suggesting….

The logo

The logo is kept per product and used by every campaign of that product. Click Upload logo and pick a PNG, JPEG or WebP file, up to 20 MB. A transparent PNG works best on any background.

The template places it above the headline (top left in Overlay’s square and tall sizes), up to 7.5% of the ad’s shorter side high. Creatives already rendered keep their old images: click Render again on a creative to add the new logo. Remove deletes the logo; creatives keep it until they’re rendered again.

What drawing costs

Each creative’s card and the Sizes card show the estimate for one picture, for example “about $0.05 at medium quality”, based on the image task’s model and quality in Settings → General → AI. The default is GPT Image 2.5 Flare at medium. Each picture is drawn once at 2048×2048, and every size is cropped from it, so the number of sizes doesn’t change the cost.

After drawing, the card shows what the picture cost, and the Sizes card shows what this campaign’s pictures have cost so far. It all counts toward your monthly AI budget. Quality and prices are in Costs: Ad pictures.

The rules every picture is drawn under

The prompt sent to the image model starts with fixed rules, then your product’s name and one-line description, then what you wrote:

• no text, letters, numbers, captions, logos, watermarks or screens with readable words, because the words are added later;
• no real or identifiable people, celebrities or public figures; people, if any, are generic and not a face close-up;
• no brand names, trademarks, or other companies’ products, packaging or app icons;
• no before-and-after comparisons, body-image or health claims, or anything shocking;
• the subject near the center with generous margins, so it can be cropped to wide and tall sizes;
• for Overlay, a calm lower third where the words sit; otherwise a clean background;
• natural light and a modern, credible look, not stock-photo clichés.

The full prompt is kept with the creative: open Prompt sent to the image model on its card. If OpenAI declines a prompt, the creative shows “OpenAI declined to draw this” with OpenAI’s reason. Change what the picture shows and draw again.

Rendering

Every size is rendered on your computer, not by the AI. Startica writes the template as a small web page, renders it to an image in a hidden window at exactly the ad’s pixel size, and saves it. Only the template, the picture and the logo can load in that window: nothing is fetched from the internet. See Privacy and security.

Each size is saved as a PNG, which keeps text sharpest. If the PNG is over the platform’s file limit, it’s saved as a JPEG instead, at the highest quality that fits (92, 85 or 75). If even that’s too big, rendering fails with “The width×height image is bigger than the platform takes”.

Rendering needs the desktop app. In browser development mode it uses Playwright’s Chromium if it’s installed; otherwise it fails with “Images can’t be rendered here: open Startica’s desktop app (or install Playwright’s Chromium for browser dev mode)”.

Work with a creative

Each card shows the headline, the status, the template, which ad it’s for and what its picture cost. Below the preview (the 4:5 size, or 1:1) are the picture’s description, a download button per size with its file size, and the buttons:

Button	What it does
Edit	Opens the dialog. Keep the current picture is selected, so saving only renders again. Pick Draw a new one with AI or Upload my own to replace the picture.
Draw again	Draws a new picture from the same description, after you confirm “Draw a new picture? It costs about $…”. Only for AI pictures, with an OpenAI key.
Render again (the arrows icon)	Renders every size again, for example after changing the logo or after a failed render.
Stop	While drawing or rendering. The creative keeps its earlier images, if it had any, and shows “Stopped by you”.
Trash icon	Deletes the creative and its images, after you confirm.

Status	Meaning
Idea	Suggested by the AI; no picture yet.
Drawing	The AI is drawing the picture.
Rendering	The sizes are being rendered.
Ready	Every size is ready to download.
Failed	Drawing or rendering didn’t finish. The error is shown on the card.

A creative can’t be edited while it’s drawing or rendering (“Wait for this creative to finish”).

Download and use

Click a size to download it. Files are named after the creative and the size, for example creative-7b1e4c2a-4x5.png. Then add them to the ad in the platform:

• Meta: after the bulk import, add the 4:5 (or 1:1) image to each ad in Ads Manager, and the 9:16 for Stories and Reels placements.
• Reddit: add the 4:5 or 1:1 image to the promoted post when you create the ads.
• Google: use them in a Demand Gen, Display or Performance Max campaign. Search ads don’t take images.

Deleting an ad in Startica keeps its creatives; they become Whole campaign. Deleting a campaign deletes its creatives and their files.

Where the files are

Pictures, rendered sizes and logos are files in the data folder: ads/creatives/<creative id>/ and ads/logos/. Uploads stay there and are never sent anywhere.

In the demo

Demo mode gives the Reddit campaign two creatives: “Your field notebook, minus the soggy pages”, rendered in Overlay from a stand-in picture drawn on your computer (it stays an Idea if the demo can’t render), and “Stop typing up field notes at night”, an idea in Panel not drawn yet. No background jobs run in the demo, so Suggest ideas, drawing a picture and rendering (after a save or Render again) all queue a job that just waits. You can download the rendered sizes.

Related

• Ads
• Results and what to do next
• Hand-off
• How Startica uses AI
• Costs