Prompt Studio — build it, don't describe it.
A two-stage builder for branded graphics: fill in a short brief, pick a layout from real example renders, and get back a six-section image prompt you can edit line by line.
Describing a good graphic in a sentence is hard. Prompt Studio replaces the sentence with a short form and a carousel of real example renders — you fill in a headline, pick a look, and it writes the production-grade image prompt for you, in six labelled sections you can edit.
What it is
A two-stage builder. Stage 1 is the brief: what the graphic says, what's in it, and what your brand's rules are. Stage 2 is the prompt it writes from that brief — structured, sectioned, and fully editable before you use it.
It exists because the hard part of an on-brand AI image isn't the render. It's the prompt: the palette in a sensible ratio, the fonts, the text reproduced verbatim so the model doesn't invent its own words, a layout named rather than described, and a list of things to avoid. Prompt Studio writes all of that.
Three ways in
- From Compose. Under any text reply, hit Make graphic. It reads the post, pulls a headline out of it, and opens the Studio with the brief already filled in. It also switches the composer to image mode at 1:1.
- From your Daily Brief. Every drafted social card has a Prompt Studio action. Same builder, and its primary button is Create in Compose — you land back in the composer with the prompt loaded.
- From BlogCraft. The Visuals menu in the editor opens it for a blog thumbnail or an infographic, seeded from the finished article.
The format decides the shape, and the shape decides the aspect ratio — you don't set it by hand:
| Format | Aspect | Comes from |
|---|---|---|
| Social post graphic | 1:1 | Compose, or a brief's social card |
| Blog featured thumbnail | 16:9 | BlogCraft → Visuals |
| Blog infographic | 4:5 | BlogCraft → Visuals |
Stage 1 — the brief
Headline — the only required field
Up to 80 characters. This is the text that will physically appear in the image, large and bold. It's reproduced verbatim in the prompt — the model is explicitly told not to reword it. Nothing else on this page is mandatory; without a headline, Generate stays disabled.
Sub heading
Up to 90 characters. The smaller line under the headline. Leave it blank and the prompt simply doesn't mention one.
Kicker — blog thumbnails only
Up to 48 characters. The little eyebrow label above the headline (“GUIDE”, “CASE STUDY”). This field only appears on the blog-thumbnail format.
Topics — up to five key points
90 characters each, five maximum. Expanded by default on infographics, where they're the whole point; collapsed behind an Add topics button everywhere else. Each one is rendered in the image as a bullet.
Scene
Up to 200 characters. What is actually depicted — “an engineer's van parked outside a terraced house at dawn”. Leave it empty and you get a clean, premium graphic with a single focal subject and no scene of its own.
Brand Rules — auto-filled from your Visual Identity
Up to 400 characters, and already written for you. It's built from the brand's real palette, real fonts and logo: your hex codes in a 60/30/10 split, your typefaces, and a clean corner reserved for the logo. Edit it if you want, but the default is the right answer for most graphics.
Brand Rules is only as good as Brand → Visual Identity. If your palette was extracted wrong, every graphic you ever build inherits the wrong colours. Two minutes there fixes it everywhere.
The layout carousel
Layout sits behind an Auto row at the bottom of the builder. Open it and you get a paginated carousel of real example renders — not wireframes, not icons. You're picking a picture, not describing one.
- Social — 20 layouts. Centered, Lower-left, Top banner, Quote card, Big stat, Split, Diagonal, Magazine, and a dozen more.
- Blog thumbnail — 10 layouts. Lower-left, Centre band, Left panel, Split 50/50, Framed, and so on.
- Blog infographic — 10 layouts. Cards, Timeline, Numbered rows, Zig-zag, Circular, Pyramid, Comparison, Stat grid, and more.
Auto is a real choice, not a placeholder. Leave it on Auto and the Studio keeps the layout it picked for you. Pick one from the carousel and it overrides.
Stage 2 — the prompt, in six sections
Hit Generate prompt and the builder hands off to an AI art director, which turns your brief into a structured prompt with six labelled sections. Each one lands in its own editable card:
SUBJECT & SCENE // one vivid sentence, one focal subject IN-IMAGE TEXT // your headline, sub heading and topics — verbatim LAYOUT & COMPOSITION // the named layout, described BRAND RULES // palette, fonts, logo space STYLE & MOOD // four to eight words AVOID // the pitfalls: garbled text, clip-art, wordmarks
You can rewrite any section in place — tighten the scene, add a colour, delete a bullet from the in-image text. The prompt is reassembled from whatever the cards say when you leave.
Three buttons on this stage:
- Back to builder — returns to the form with everything you typed intact.
- Copy prompt — the edited prompt to your clipboard, ready for Midjourney, DALL·E, Flux or Ideogram.
- The primary button — Apply to prompt in Compose, Create in Compose from a brief. Either way, the prompt lands in the composer at the right aspect ratio.
Half of what makes an AI graphic look amateur is what's in it that shouldn't be: misspelled in-image text, a hallucinated wordmark, stock-photo cliché. The AVOID section names those explicitly. If you delete it, expect worse results.
What it costs
- Generate prompt: 2 credits — and only when the AI pass succeeds.
- The built-in structured fallback is free. If the AI pass fails or comes back malformed, ScoutRival assembles the same six sections itself, deterministically, from your brief. You still get a complete, well-formed, brand-locked prompt — you just aren't charged for it. You won't be told which one you got, and for most briefs you won't be able to tell.
- Everything else in the Studio is free — opening it, editing the brief, switching layouts, editing the sections, copying the prompt.
Rendering the prompt into an actual picture is a separate step, on your own connected image model, and costs 0 credits. Image mode →
What to do with the prompt
- Render it in ScoutRival. Connect an image model under AI Models and the reply gets a Generate with my model button. The picture renders on your provider's account and lands in the thread.
- Or paste it anywhere. Copy it and use whatever image tool you already pay for. The prompt is written to be portable.