Anjin Design
Start free

Render API

One call.
One on-brand image.

No headless browser. No font-loading race. No screenshot service quietly rendering the wrong weight.

Start free →Read the docs
curl -X POST https://api.anjin.design/v1/renders \
  -H "Authorization: Bearer $ANJIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template": "acme_instagram_1080x1350",
    "modifications": [
      {"name": "headline",    "text": "Launch day"},
      {"name": "hero_image",  "image_url": "https://…/photo.jpg"},
      {"name": "brand_color", "color": "#2563EB"}
    ]
  }'

Why the output is trustworthy

One engine, two homes.

The renderer in the browser editor and the renderer behind this API are the same engine running the same file. Parity is structural, not a test suite we hope stays green. We measure it at SSIM 1.00000.

If you have ever shipped an image pipeline on headless Chrome, you know the class of bug this deletes: the font that resolved differently on the render box, the line that wrapped one word early in production, the emoji that came out a box.

Structural similarity

1.00000

Editor output vs API output. A perfect match.

The surface

Seven endpoints. That’s the whole thing.

GET /v1/templatesPublished templates, each with its addressing key
GET /v1/templates/:id/schemaThe slot contract — names and types
POST /v1/rendersRender one. Async by default, or wait for the image
POST /v1/renders/batchRender many, with per-row success and failure
GET /v1/renders/:idPoll a job
POST /v1/assets/querySearch your asset library by tag or metadata
GET /v1/webhook-secretYour HMAC secret for verifying callbacks

Discover before you render

The template tells you what it accepts.

Validate input before you spend a credit — and keep your integration working when a designer adds a slot. An unknown slot name fails with a 422 naming the offender, and charges nothing.

GET /v1/templates/:id/schema

{
  "slots": [
    { "name": "headline",    "type": "text"  },
    { "name": "hero_image",  "type": "image" },
    { "name": "brand_color", "type": "color" }
  ]
}

Two URLs, on purpose

One to download. One to keep.

output_url

A signed link, valid about an hour. Fine for an immediate download.

asset_url

Permanent. Use this for anything you store or hand to another system.

We learned this the hard way: a temporary URL saved into someone’s database expires, and the failure surfaces weeks later as a broken image on a client’s site.

Formats
NORTHWIND

Launch day

1080×1350

NORTHWIND

Launch day

PRODUCT

1080×1920

NORTHWIND

Launch day

PRODUCT

1080×1080

NORTHWIND

Launch day

PRODUCT

1280×720

Each one composed for its shape — not the same image cropped four ways.

Errors worth knowing

Structured, and free when they’re our fault.

422template_not_publishedPublish the frame first. No credit charged.
422unknown_slotWith the offending slot name. No credit charged.
422image_out_of_folderA folder-bound slot was given an outside image.
402insufficient_creditsWith the amount required.
400invalid_scalescale must be between 1 and 4.

Failed renders are refunded automatically. You’re charged for images you receive.

Cost

One credit. More only when it’s more work.

scale 1–4×, priced scale² — 1, 4, 9, 16 — because that is what the pixels do. The response tells you the cost.

Usage
northwind_instagram_1080x1350API1
northwind_story_1080x1920API1
kepler_thumbnail_1280x720Dashboard1
halcyon_hero_1920x1080API · 2×4

A 2× render costs 4 credits — four times the pixels, four times the price.

Also available

Batch

Send many, get per-row success and failure. One bad row fails alone; the rest proceed.

Signed webhooks

Completion callbacks signed with X-Anjin-Signature. Fetch the secret from the API and verify it.

CLI + MCP

anjin render … for shells and CI, and an MCP server so AI agents can render on brand.

Questions

How do I generate images from a template with an API?
POST to /v1/renders with a template key and your content as typed slots. You get back a job with a signed output_url and a permanent asset_url. There is a batch endpoint for volume and signed webhooks for completion.
Will the API output match my design exactly?
Yes. The editor and the API run the same rendering engine on the same file, so parity is structural rather than tested for. We measure it at SSIM 1.00000.
Is there a free image generation API?
The free trial gives you 30 credits with no card, which is enough to build a template and render from the API properly. Paid plans start at £29 a month.
Can I generate images in bulk?
Yes — POST /v1/renders/batch returns per-row success and failure, so one bad row fails alone and the rest of the batch proceeds.
What image formats and sizes are supported?
PNG, JPG and WebP at the template's own dimensions, with a scale multiplier from 1× to 4×. Each output size is a separate frame with its own layout, so it is composed rather than cropped.
Do the image URLs expire?
output_url is signed and valid about an hour. asset_url is permanent — use that one for anything you store.

Render your first image in five minutes.

Start free →

30 credits. No card. No watermark.