Image mode — how it really works.
Compose writes you a brand-aware image prompt, not a picture. Here's why, how to render one in-app with your own key, and what each step costs.
Switch Compose to Image and you get a brand-aware image prompt — not a rendered picture. That's a deliberate design decision, and this page explains it and what to do with it.
What image mode does
Describe what you want. ScoutRival writes you a production-ready prompt that already carries:
- Your brand palette — the actual hex codes from your Visual Identity, in a sensible 60/30/10 split
- Your typography — your heading and body fonts
- Composition and layout — where the text sits, where to leave room for your logo
- A negative prompt — the things to avoid, which is half of what makes AI images look amateur
You get it back with a Copy button. Paste it into Midjourney, DALL·E, Flux, Ideogram, or whatever you already use.
Why a prompt and not a picture
Rendering images is the single most expensive thing an AI product can do, and image models change every few weeks. Rather than bake in one model, mark it up, and charge you for it, we do the part that's genuinely hard — turning your brand into a prompt that produces on-brand output — and let you render on whatever you like, at cost.
If you already pay for Midjourney, you don't pay us for images too. If you don't, connect a key and render for cents.
Getting an actual image inside ScoutRival
You can render in-app. It takes one setup step.
Connect an image-capable provider
Go to AI Models and connect OpenAI, Google (Gemini), or OpenRouter. Note that Anthropic can't generate images — it's text only.
Route the Image row to it
On the same page, the routing section has three rows: Text, Blog and Image. Pick your connection and a model on the Image row.
Hit “Generate with my model”
Back in Compose, that button now appears under every image prompt. The picture renders on your provider's account and lands in the thread — with a lightbox, a download, and a regenerate.
Text and Blog fall back to ScoutRival's included AI when you haven't connected anything. Image doesn't. With no image model connected, you will only ever get the prompt — the button will say “Connect an image model to render” instead. That's not a bug.
Prompt Studio — the better way in
Typing a description into the box works. But for anything you'll actually publish, use Prompt Studio.
It's a two-stage builder. First you fill in a short brief — headline, subheading, key points, the scene, and your brand rules (pre-filled from your Visual Identity). You pick a layout from a carousel of real example renders rather than trying to describe one in words. Then it writes a structured, six-section prompt, which you can edit section by section before using.
You'll reach it from three places: the Graphic Studio chip in Compose, Prompt Studio on any social card in your Daily Brief, and Visuals in the BlogCraft editor.
Aspect ratios
- 1:1 — square, for the feed
- 4:5 — portrait, Instagram's best-performing shape
- 3:4 — portrait, Pinterest
- 16:9 — wide, for a blog hero or a YouTube thumbnail
- 9:16 — stories and reels
When you come in from BlogCraft or a brief card, the ratio is already set to the right one for the job.
What it costs
- Writing the image prompt: 2 credits. Image mode is a Compose message, and it's billed like one — you're paying for the prompt, which is the work.
- Prompt Studio's “Generate prompt”: 2 credits, and only when the AI pass succeeds. If it falls back to the built-in structured prompt, that's free.
- Rendering with your own key: 0 credits. You pay your provider directly.
Image mode costs 2 credits even though you don't get a picture back. You're being charged for the prompt, not the render. If that isn't what you wanted, connect an image model — then the render itself is free to you.
Troubleshooting
- “Where's my image? I only got text.” That's image mode working as designed. Connect an image model under AI Models to render in-app.
- The Quality picker doesn't seem to do anything. It doesn't. It's a leftover from when we rendered images ourselves, and we're removing it.
- I can only generate one image at a time. Correct — one per message, by design.
- “Your image model couldn't render this.” Check the model you picked on the Image row actually supports generation. On OpenRouter, image support is reliable on Google's models; several other listed image models aren't fully wired up yet.