Layerize Text API

Overview

The Layerize Text endpoint detects text regions in an image and returns a clean base image with all text removed. Pass any image with visible text: your own branded assets, marketing materials, generated designs, or uploaded artwork. The endpoint handles detection and removal in one call.

1 Upload an image

→

2 Get text-free base

→

3 Overlay your own text

Endpoint reference

Endpoint	Method	Description
`/v1/ideogram-v3/layerize-text`	POST	Extract text layers from an image

Request

Send the image as multipart/form-data.

Parameter	Type	Required	Description
`image`	binary	Yes	The image to analyze. JPEG, PNG, or WebP (max 10 MB).
`prompt`	string	No	Description of the image. Improves detection accuracy. Auto-generated if omitted.
`seed`	integer	No	Random seed for reproducible results.

Response

Field	Type	Description
`base_image_url`	string	URL of the image with all detected text removed.
`original_image_url`	string \| null	URL of the original image with text intact.
`seed`	integer	Seed used for this layerization.

Layerize an image

Upload any image that contains text. This can be a branded poster, a marketing email screenshot, product packaging, or anything with visible typography.

Format: multipart/form-data

Max size: 10 MB per image

Accepted: JPEG, PNG, WebP

import requests

# Upload your image directly
with open("poster.png", "rb") as f:
  response = requests.post(
    "https://api.ideogram.ai/v1/ideogram-v3/layerize-text",
    headers={"Api-Key": "<apiKey>"},
    files={"image": ("poster.png", f, "image/png")},
    data={"prompt": "A concert poster with event details"}
  )

layers = response.json()
print(f"Base image (text removed): {layers['base_image_url']}")

curl -X POST https://api.ideogram.ai/v1/ideogram-v3/layerize-text \
  -H "Api-Key: <apiKey>" \
  -F [email protected] \
  -F prompt="A concert poster with event details"

Use the output

The response gives you a base image with text cleanly removed and the original for reference. Use the base image as a background and overlay your own text with any font, size, or language. Useful for localization, A/B testing copy, or repurposing designs across campaigns.

# Example response
{
  "base_image_url": "https://ideogram.ai/assets/image/...",
  "original_image_url": "https://ideogram.ai/assets/image/...",
  "seed": 42
}

# Download the base image for compositing
base_img = requests.get(layers["base_image_url"]).content
with open("base_no_text.png", "wb") as f:
  f.write(base_img)

# Now overlay your own text using Pillow, Cairo, or any compositing tool

Combine with Generate

You can also generate a design first and then layerize it. This is useful when you want Ideogram to handle the visual design and you need to swap out the copy afterward.

# Generate a design
gen = requests.post(
  "https://api.ideogram.ai/v1/ideogram-v3/generate",
  headers={"Api-Key": "<apiKey>"},
  data={
    "prompt": "A travel poster with the text 'Visit Tokyo'",
    "aspect_ratio": "2:3"
  }
).json()

# Download and layerize
img_data = requests.get(gen["data"][0]["url"]).content
layers = requests.post(
  "https://api.ideogram.ai/v1/ideogram-v3/layerize-text",
  headers={"Api-Key": "<apiKey>"},
  files={"image": ("design.png", img_data, "image/png")}
).json()

# base_image_url now has the design without text
print(layers["base_image_url"])

Tips

Works on any image with text. Uploaded photos, screenshots, branded assets, AI-generated designs. The image does not need to come from Ideogram.
Pass a prompt to improve detection accuracy. A short description of the image helps the model distinguish text from visual elements.
Set a seed for reproducible results. The same image + seed will produce the same output.
Use for localization. Layerize once, then overlay translated copy in any language without regenerating the design.

Start layerizing text via the API.

Try it in Ideogram Back to Text Layers