Inpaint


Mask the sky, type “soft clouds at sunrise”, and only the masked area is repainted. The mask must match the picture; the editor guarantees it.
AI image studio
Opalframe is a self-hosted studio: prompts and styles, a canvas editor with masks and outpainting, background removal, upscaling, batches, brand kits, boards and credits. Plug in your provider, or run the built-in simulated engine.




The studio
Pick a style, a ratio and a seed. The same seed and settings give the same picture, so a result you like can be reproduced, remixed or varied.
Prompt
trail backpack on a mossy rock at sunrise
Style
4 creditsDraft




Sample art from the built-in simulated renderer. Progress shows per picture while the worker renders.
Style presets
Presets add a prompt suffix and a negative prompt, and the swatch tells you what you will get. Operators can edit them.

Watercolour

Flat vector

Risograph

Ink

Gouache

Paper cut

Pastel

Neon line

Mid-century

Flat block

Pixel

Charcoal
Edits
The canvas editor is plain 2D: a mask brush and eraser, zoom and pan, and an outpaint frame. The result is linked to the original as its edit, so the chain stays visible in the library.


Mask the sky, type “soft clouds at sunrise”, and only the masked area is repainted. The mask must match the picture; the editor guarantees it.


Pick a ratio and where to keep the picture. Only the new area is painted; the original pixels stay untouched.
Background and upscale
One-click tools for product shots and print sizes. Both work on library pictures and on uploads.


Transparent PNG in one step. Simulated pictures are cut from the renderer’s own subject layer; uploads use an edge heuristic.


Lanczos resize with light sharpening, capped at 8192 px. A before/after slider shows what changed. No detail is invented.
Batch and brand kits
prompt,style,ratio,seed trail backpack on a mossy rock,gouache,3:2,101 insulated bottle on a granite ledge,flat-vector,1:1,102 tent glowing under a starry sky,risograph,3:2,103












Each row is one job. A contact grid is built per request and the whole batch downloads as a ZIP. Rows that cannot be afforded are skipped, not charged.
Copperleaf Outdoor Co.



Palette colours steer the renderer and keywords are appended to every prompt. Agencies keep one kit per client.
Run it as a service
Masonry library with filters, edit chains and provenance. Boards collect campaigns; share links show only that board and expire.
Blocklists, a real-person name list, a review queue and a three-strike workspace lock. Not an AI classifier: you tune the lists.
Credits are reserved when a job starts, settled per finished picture and refunded when one fails or you cancel. Plans, packs and an agency overdraft.
Jobs run in a separate worker with leases and crash recovery. A re-queued job renders the same pixels and never charges twice.
Credits and plans
A Draft picture costs 1 credit, Standard 2 and Detail 4. Background removal and upscaling are 1 to 2. Top up with packs any time.
Try the studio
$0/mo
For solo creators
$19/mo
For teams and shops
$49/mo
For agencies with client brands
$149/mo
Credit packs: 100 credits for $9 · 500 credits for $39 · 2,000 credits for $129. Plan prices and packs are yours to set.
REST API and webhooks
Scoped API keys, an OpenAPI 3.1 reference, idempotency keys on every write, and signed webhooks for finished, failed and partial jobs.
curl https://studio.example.com/api/v1/generations \
-H "Authorization: Bearer $OPALFRAME_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "ceramic vase with eucalyptus",
"style": "ink", "ratio": "2:3", "count": 4,
"model": "standard",
"idempotencyKey": "campaign-17-vase"
}'
# 202 → { "id": "…", "status": "queued", "credits": { "reserved": 8 } }Bring your own provider
Opalframe includes a simulated renderer so you can try everything today. Real providers are adapters: off until you set a key, locked until you enable live calls.
Any endpoint that speaks the images generations and edits protocol. Set a key and a base URL.
Generate, edit, remove background and upscale through a Stability-style service.
Create a prediction, poll until it ends, collect the output. Polling is capped and pinned.
No. Opalframe ships a simulated renderer that produces procedural sample art, so you can run every feature locally. For real images you connect your own provider through an adapter (OpenAI-compatible, Stability-style or Replicate-style) with your own key and costs.
The picture was drawn by the built-in procedural engine from your prompt, style and seed. It is deterministic sample art, not an AI model, and it carries a small mark and a Simulated badge so nobody mistakes it for provider output.
Credits are reserved when a job starts, settled for every picture that finishes and refunded for the ones that fail or are cancelled. Plans grant a monthly allowance; owners can buy packs, and the Agency plan allows an overdraft.
Yes. Each workspace has a “Never send my content” switch. With it on, only the simulated engine runs for that workspace, even if the operator has connected a live provider.
Prompts are checked against operator-editable blocklists and a real-person name list before any credit is reserved. Matches go to a review queue, and three violations lock a workspace until an operator unlocks it. It is a term list, not an AI safety classifier.
Node.js, MongoDB with a replica set and the worker process. A setup wizard creates the admin account and the first workspace. There is no GPU requirement because the model, if any, lives at your provider.
Try the studio with the demo logins. Everything runs on the simulated engine until you connect a provider.