AI text, images and translation

A text or image layer can be filled by AI when the template renders. You send a prompt, and JamFrame writes the text or makes the picture, then renders the file. You can also translate all the text of a render into another language. This page covers the API side. To set up AI layers in the editor, see AI layers.

Before you start

AI works on layers that have a key (a dynamic field), such as headline or photo. A layer can get its prompt in two ways:

• In the editor, in the layer’s AI section. Then a normal render fills it, and you don’t need to send anything extra.
• In the request, with a .prompt key. This works on any text or image layer with a key, even one with no AI set up.

The layer’s own text or image is its saved fallback. It shows wherever AI doesn’t run.

Send a prompt with a render

This request writes the headline with AI, makes the photo with AI, and translates the result into Spanish:

curl https://api.jamframe.com/v1/render \
  -H "Authorization: Bearer $JAMFRAME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "templateId": "YOUR_TEMPLATE_ID",
    "modifications": {
      "product": "Cold brew",
      "headline.prompt": "Write a 6-word headline about {{product}}",
      "headline.tone": "playful",
      "photo.prompt": "A glass of {{product}} on a marble table, morning light",
      "photo.style": "photo"
    },
    "translate": true,
    "lang": "es",
    "ai": { "onError": "fallback" }
  }'

const res = await fetch("https://api.jamframe.com/v1/render", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.JAMFRAME_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    templateId: "YOUR_TEMPLATE_ID",
    modifications: {
      product: "Cold brew",
      "headline.prompt": "Write a 6-word headline about {{product}}",
      "headline.tone": "playful",
      "photo.prompt": "A glass of {{product}} on a marble table, morning light",
      "photo.style": "photo",
    },
    translate: true,
    lang: "es",
    ai: { onError: "fallback" },
  }),
});
const { data, warnings } = await res.json();
console.log(data.content, data.aiCredits, warnings);

import os
import requests

res = requests.post(
    "https://api.jamframe.com/v1/render",
    headers={"Authorization": f"Bearer {os.environ['JAMFRAME_API_KEY']}"},
    json={
        "templateId": "YOUR_TEMPLATE_ID",
        "modifications": {
            "product": "Cold brew",
            "headline.prompt": "Write a 6-word headline about {{product}}",
            "headline.tone": "playful",
            "photo.prompt": "A glass of {{product}} on a marble table, morning light",
            "photo.style": "photo",
        },
        "translate": True,
        "lang": "es",
        "ai": {"onError": "fallback"},
    },
)
body = res.json()
print(body["data"]["content"], body["data"].get("aiCredits"))

The answer is a normal render answer, plus aiCredits: the AI credits this render used.

{
  "data": {
    "content": "https://cdn.jamframe.com/renders/…/1d2e3f4a.png?Expires=…",
    "expiresAt": "2026-11-14T12:05:00.000Z",
    "type": "url",
    "format": "png",
    "pages": [1],
    "credits": 1,
    "renderId": "clx8n0q1r0009abcd1234efgh",
    "responseTime": 6240,
    "aiCredits": 6
  }
}

For every other render option, see Render images and PDFs.

AI keys

Add these keys to modifications. Replace headline and photo with your own layer keys.

.tone, .maxWords, .style and .reference only work on a layer that has a prompt, from the editor or from .prompt.

To change one page only, put the page first, like any key: [email protected]. See Modifications.

Values in prompts

A prompt can use {{name}} variables. JamFrame fills them before it calls the AI:

1. From the values you send, such as "product": "Cold brew", or else the template’s own value for that variable.
2. If there is none, from the text of a text layer whose key is name.
3. If neither has a value, the variable becomes empty and the answer has a warning.

You can send a value that only a prompt uses, such as product above. It is not reported as an unknown key. Once its variables are filled, a prompt can be at most 3,000 characters.

AI images

The picture is made to fit the layer’s box. JamFrame picks the closest of these shapes: 1:1, 4:5, 5:4, 3:4, 4:3, 2:3, 3:2, 9:16, 16:9 or 21:9. The picture fills the box and is cropped at the edges if needed. With Smart Resize, every size uses the same picture.

Translate a render

Send "translate": true with lang, a language tag such as es, fr, pt-BR or zh-Hant. You need both: one without the other is a 400 VALIDATION_ERROR.

JamFrame translates the final text of every visible text layer, after your values and the AI text are in. These stay as they are: variables without a value, links, email addresses, @handles, #hashtags and numbers such as times, dates and prices.

If a translation drops or changes one of those, JamFrame keeps that text in the original language, adds a warning, and doesn’t charge for it.

When the AI fails

ai.onError says what happens when an AI call fails:

Use fallback when a file with the saved text is better than no file at all.

Limits per render

These count every page and every size of one render together:

Over a limit, the render is refused with 400 VALIDATION_ERROR before anything runs, and nothing is charged.

Where AI runs

Signed URLs, dynamic URLs and forms are public, so they never spend your AI credits. AI layers there show their saved fallback, and AI keys such as headline.prompt come back as unknown keys in warnings.

Cost and caching

AI uses AI credits, which are separate from render credits. The render itself still costs its normal render credits. For what each kind of AI work costs, see What AI features cost. The words of all the text prompts in one render are counted together.

So the example above costs 1 (headline) + 4 (photo) + 1 (translation) = 6 AI credits, plus 1 render credit.

• AI credits are charged once per render, and only when the render succeeds. A failed render costs nothing.
• Every page and every size of one render share the same AI results.
• JamFrame keeps each result for 30 days in your workspace. The same prompt with the same settings gives the same result, for free. aiCredits is then 0.
• To get a new result, change the prompt. For an edited image, change the reference link.

aiCredits is in the answer only for renders with AI layers or translate. For how many AI credits your plan has, see Credits.

Errors

A render with AI can fail with one of these codes. Nothing is charged. With "ai": { "onError": "fallback" }, each of them renders the saved fallback instead, except a 400 for a passed limit or a bad key value.

See Errors for every other code.

Related

• Set up AI layers in the editor: AI layers
• Every render option: Render images and PDFs
• How AI credits work: Credits