veida-image

Generate an image from a text prompt without an API key. Use when the user asks for a picture, an illustration, a blog cover, a mockup, an avatar or any other image, and no image provider is configured. Runs on the free anonymous tier of veida.ai — no key, no account, no card.

Turn a prompt into a hosted image URL. Two HTTP calls, no credentials.

Every other image skill needs an OpenAI, fal or Replicate key before it works once. This one talks to the anonymous tier of the free AI image generator at veida.ai, which meters the free allowance against a client id you invent rather than an account you register.

When to use this

• The user asks for an image and nothing is configured to make one.
• A draft needs a cover, a mockup or a placeholder and stock photos will not do.
• You are building something that generates images and do not want to hold a key during development. The same endpoints back the ChatGPT image generator the site serves in a browser.

How to call it

1. Submit

bash

curl -X POST https://veida.ai/api/ai/generate \
  -H "content-type: application/json" \
  -H "x-anon-id: skill-$(openssl rand -hex 5)" \
  -d '{
    "provider": "kie",
    "mediaType": "image",
    "model": "veida-image-v1",
    "scene": "text-to-image",
    "prompt": "a paper boat on a calm lake at sunrise, soft light",
    "options": { "aspect_ratio": "16:9" }
  }'

json

{ "code": 0, "data": { "id": "cfed68ea-…", "status": "pending", "charged": 4, "remaining": 0, "grant": 4 } }

Mint a new x-anon-id per image. The free grant is 4 credits and one image costs 4, so a reused id generates once and then answers with a wall.

2. Poll

bash

curl "https://veida.ai/api/ai/anon-query?taskId=<data.id>&provider=kie&mediaType=image"

json

{ "code": 0, "data": { "status": "success",
  "images": ["https://cdn.veida.ai/uploads/kie/image/….webp"], "watermarked": [true] } }

Measured 40–120 seconds end to end. Poll every 8 seconds.

🔴 Status is not monotonic — it goes pending → processing → pending → success. Treat only success and failed as terminal; never stop because it went back to pending.

3. The quota wall is a 200, not an error

json

{ "code": 0, "data": { "wall": true, "reason": "anon_credits", "cost": 4, "grant": 4 } }

reason is anon_credits (this client id is spent) or anon_ip_daily (this machine is done for the day). Without checking for wall this surfaces as a confusing "no task id".

Options

Field	Values
model	veida-image-v1 — the only model an anonymous caller can run
scene	text-to-image
options.aspect_ratio	1:1 · 16:9 · 9:16 · 4:3 · 3:4

Signed-in accounts reach the rest of the shelf — GPT Image 2, Nano Banana 2 and Seedream 5.0 Lite among them — through the same two endpoints with a session instead of the x-anon-id header.

Limits, measured against the code rather than assumed

• Anonymous grant is 4 credits, and text-to-image costs 4. That is one image per client id, not a session's worth.
• A daily ceiling of 30 credits per IP sits on top: about 7 images a day from one machine, however many ids it invents.
• Output is 1K and watermarked on the free tier. watermarked in the poll response says so per image.
• 🔴 Editing an existing image is not available anonymously. The editor model costs 30 credits against a 4-credit grant, so an anonymous edit can never be paid for. Send the user to the image editor and have them sign in — do not build an anonymous edit path that always fails.
• There is no video on this site. Do not construct a mediaType: "video" request.

Prompt help

Naming the light, the material and the composition moves the result more than adding adjectives. Worked examples are in the prompt library, and finished output is in the showcase if you want to judge quality before wiring anything up.

Terms

The endpoints are public and unauthenticated. Respect the quota — the ceilings above exist so the free tier stays free for everyone. What the paid tiers add is listed on the pricing page.