> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.roor.ai/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.roor.ai/_mcp/server. # Image The Image tool makes one to four stills from a prompt. It shares its console with Video: model row, prompt, references, settings, Generate. What differs is what the models can do. ## The prompt Describe the picture: subject, setting, light, lens, mood. Type `@` to mention a saved Element (a character, a location, a prop or a style) so it appears the way you defined it. The prompt holds up to 5,000 characters. Two helpers sit under the prompt: * **AI writer** rewrites your prompt into something a model responds to better. It costs no credits. **Undo** brings your words back. * **History** reopens prompts you have used before, and offers a few starters when you have none yet. There is no negative prompt for images. Put what you do not want in plain words in the prompt instead. ## References Press **Add references** to attach images that steer the result. Pick from **Uploads**, **Assets**, **Elements** or your earlier **Takes**. Each model sets its own limit; the picker counts them off for you. * **Editing a photo.** On models that edit, attach the photo and say what should change: "make it night", "remove the sign". The console tells you when the selected model can do this. * **Style from a reference.** Some models take their whole look from one reference image and will not run without it. * **Your own fonts.** Typography models accept up to two font files (TTF, OTF, WOFF or WOFF2). Put the words to set in quotes: `"ROOR STUDIO"`. Each font adds a little to the price. * **Paid references.** A few models charge for every reference they read. When that applies, the console says so and the Generate price already includes it. ## Style The **Style** panel adds direction without you writing it: **Genre**, **Color palette**, **Lighting** and **Lens & framing**. Your picks are turned into language and appended to the prompt when you generate. ## Settings | Control | What it sets | | ---------- | ------------------------------------------------------------------------------------------------ | | Quality | 1K, 2K or 4K. Free and Starter render at 1K; Creator and above unlock 2K and 4K. | | Aspect | 1:1, 16:9, 9:16, 4:3, 3:4 or 21:9, as the model allows. | | Count | One to four images from one prompt. Each is its own take, priced and refunded on its own. | | Background | Full frame, or cut out with a transparent background, on models that support it. | | Seed | Under Advanced on models that support it. Lock it to reproduce a result; reroll for a variation. | Changing the model re-checks your settings. If the new model does not offer your aspect ratio or quality, the console resets it and tells you. ## Models The picker lists what is available to you, cheapest first, with a one-line summary of each model's strengths and its starting price. The Seedream family is the studio's core: | Model | Credits per image | References | Notes | | ----------------- | ------------------------- | ---------- | ----------------------------------------------- | | Seedream 4.0 | 25 at any size | up to 10 | Recommended. Photoreal, strong prompt fidelity. | | Seedream 4.5 | 35 | up to 10 | Edits photos. | | Seedream 5.0 Lite | 35 at 1K, 30 at 2K and 4K | up to 10 | | | Seedream 5.0 Pro | 40 at 1K, 80 at 2K and 4K | up to 10 | Best quality; large sizes cost more. | More models appear in the picker as they are switched on, each with its own price shown before you generate: the Nano Banana and GPT Image families for prompt-faithful illustration and edits, FLUX.2 for photoreal detail, Recraft for vector and brand-style work (SVG output on the Vector models), Riverflow for typography with your own fonts, Qwen Image, Grok Imagine and MAI Image. The summary line under each name says what it does: edits photos, cut-out backgrounds, your fonts, and so on. ## Generate The **Generate** button shows the price. Press it, or use Ctrl+Enter (Cmd+Enter on a Mac). Press `/` to jump to the prompt. In the rail, `G` and `L` switch between grid and list. If the prompt is empty the button nudges you to describe the shot. If you are short of credits, the top-up sheet opens instead. ## A finished image Open a take from the rail: * **Animate** sends the image to Video as the start frame of a clip. * **Reuse** puts the prompt, model, settings and references back in the console. * **Regenerate** in the rail row renders the same prompt again, at the price shown. * **Download**, **Favourite**, **Copy URL** (while the take is private). * **Create element** turns the image into a reusable Element, or assigns it to one you have. * **Send to Community** submits it for review. See [Community](/share/community). * **Move to project** files it under a project. * **Delete**, with an inline confirm. The details panel shows the model, quality, size, date, cost, and a **Take** code. Quote that code if you write to support about a take. ## When something goes wrong * **A refused prompt** shows "Inappropriate prompt detected. Generation blocked." under the prompt. Nothing is charged. See [Content rules](/help/content-rules). * **A failed render** shows FAILED with a plain reason, and either "REFUNDED" with the amount or "NO CREDITS WERE CHARGED". **Retry** renders a fresh take at the price shown; **Edit and retry** takes you back to the console; **Change reference** appears when a reference was the problem. * **Some of several takes did not start.** The toast says how many, and nothing is charged for those. * **Connection lost.** The console keeps your draft for 24 hours. Retry when you are back online.