On-brand images, generated from your own site.
The Spider decodes your brand from your crawled homepage — five color swatches, fonts, mood — and every cover, people scene, and product shot it generates carries them. Your exact title is composited on top, never left to the model.
Brand kit auto-extracted from your homepageTitle edits re-composite free — no re-render
From $14/mo · 7 days free · No card required
Emergency Dentist Pricing: What You'll Actually Pay
Visit fees, insurance vs cash, and financing — explained
Rendered once · title edits re-composite free
Decode, direct, render, compose
Generic AI images scream stock. The Studio's pipeline keeps your brand in charge at every step — and keeps the text out of the model's hands.
Your brand, from your site
One pass over your crawled homepage screenshot extracts five color swatches, heading and body fonts, and mood words. Edit any of it — it's your kit.
Pick a profile
Blog cover, people scene, or product shot — each with its own prompt recipe, text policy, and reference-image fidelity mode.
A scene, not a poster
The model paints a brand-colored scene with clean negative space — and is strictly told to render no text, so nothing comes back garbled.
Your title, your hexes
The exact headline and description are composited programmatically in your brand colors. Change the title later — it re-composites from the archived render, free.
Built for brands, not one-off pictures
Everything here compounds: the Studio remembers your direction, keeps every take, and never wastes a render.
An editable brand kit
Swatches, fonts, and mood live on the project and feed every prompt — color words guide the scene, exact hexes style the overlay.
Reference fidelity modes
Match a style, preserve a person or logo, or keep a product true — in product mode, the printed label survives generation.
Text that's never garbled
Models are told to render no lettering; your title is composited over the scene programmatically. AI-typo covers are structurally impossible.
Style memory
Art direction you type — and preferences distilled from which takes you keep vs delete — carry into future prompts automatically.
Image SEO by default
WebP first — the lightweight format your LCP wants — with PNG and JPEG on demand, in five aspect ratios from 16:9 covers to 9:16 vertical.
Takes are never thrown away
Raw renders are archived, so you can re-roll a cover, retry an overlay, or revisit an old take without re-running the whole workflow.
Your agent art-directs over MCP
The Studio runs as a workflow — which means any MCP agent can run it, poll it, and save the result.
Give the implants post a cover that matches our brand.
run_workflowdownload_workflow_imageRan AI - Image Studio with your brand kit — a teal clinical scene, title composited in your exact hex. Saved as WebP next to the draft. Want a 9:16 variant for socials?
Tools your agent gets
run_workflowlist_image_referencesget_workflow_rundownload_workflow_imageOne-call workflow recipes
- AI - Image Studiorun
- AI - Page Writer (MDX)run
Or just type /seo-create-content — one of 8 packaged slash commands the Spider ships for Claude Code.
Where the images plug in
Stop shipping stock-photo SEO
Crawl your site, let the Studio decode your brand, and generate your first on-brand cover today.
From $14/mo · 7 days free · No card required
Explore the full Digispot AI SEO SpiderCommon questions
Which image models does it use?
OpenAI (gpt-image-2) and Google Gemini's image models. Bring your own key and calls go direct to the vendor — your prompts and brand never pass through Digispot, and generations aren't metered by us, so you can produce as many covers as your own account allows for just your vendor bill. Prefer no setup? Use Digispot managed credits on any paid plan instead. Either way image generation needs a connected model — there's no mock mode pretending otherwise.
How does it know what my brand looks like?
From your own site. After a crawl, the Studio decodes your homepage screenshot once into a brand kit: five hex swatches, your heading and body font families, and mood words derived from the palette. It's stored on the project and fully editable — and with your explicit consent it can also be added to your Knowledge base so text generation uses the same identity.
Why is the text on covers never garbled?
Because the model never draws it. Renders are generated under a strict no-text directive — the model produces a scene with clean negative space — and your exact title and description are composited programmatically over the image in your brand hexes. Editing the title later re-composites from the archived raw render: no new generation, no cost.
What sizes and formats can I download?
Five aspect ratios chosen at generation time — 16:9 for blog covers and og:image slots, 1:1 and 9:16 for social, 4:3 and 3:4 for layouts — and any of WebP (default), PNG, or JPEG at download, converted on the fly from a single archived original. WebP keeps hero-image weight down, which is exactly where LCP problems live. Past takes are kept, so regenerating never loses the versions you liked.