Skip to main content

Ideogram 4.5

Generate images with accurate text, or edit an existing image with a text instruction guided by up to four references.
Ideogram 4.5 generates and edits images, with Ideogram’s strength in rendering text. One endpoint covers both: send a prompt alone to generate an image, or add images to edit the first one, using the rest as references. It renders at about 4 megapixels in 13 aspect ratios, or follows the input image with auto.

​
Key capabilities

  • Generation and editing in one endpoint: Text-to-image, or edit an image with up to 5 images
  • Refer to images by position: image 1 is the image being edited and image 2 onwards are the references, so the prompt can say “put the logo from image 2 on the bottle in image 1”
  • Text rendering: Headlines, logos and labels written exactly as quoted in the prompt
  • Aspect ratios: 13 ratios from 3:1 to 1:3 at about 4 megapixels, plus auto (default) to follow the input image
  • Four quality tiers: quality very_low, low, medium or high. It is the only parameter that changes the price
  • Magic Prompt and seed control
  • Prompt range: 1 to 10,000 characters
  • Async processing: Webhook notification or polling for task completion

​
Use cases

  • Text edits: Change the copy of a sign, label or poster while keeping the rest of the image
  • Brand assets: Apply a logo or product from a reference to a new design
  • Restyling: Turn a photo into an illustration, watercolor or poster
  • Posters and social graphics: Designs with legible text, from scratch

​
API operations

Create an image with the POST endpoint, then track it with the list and status endpoints. Generation returns a task ID for async polling or webhook notification.

POST /v1/ai/text-to-image/ideogram-4-5

Create a new image generation or editing task

GET /v1/ai/text-to-image/ideogram-4-5

List all Ideogram 4.5 tasks

GET /v1/ai/text-to-image/ideogram-4-5/{task-id}

Get task status and results by ID

​
Parameters

​
Output sizes

The output is up to about 4 megapixels, shaped to the requested ratio. Measured examples:

​
Frequently Asked Questions

Ideogram 4.5 is Ideogram’s image generation and editing model. You send a prompt (1-10,000 characters), optionally with up to 5 images, and receive a task ID immediately. Poll the task endpoint or configure a webhook_url to receive the generated image URL when processing completes. Each request generates one image.
Send it as the first entry of images and describe the change in the prompt, for example “change the neon sign in image 1 so it reads CLOSED in blue neon; keep everything else unchanged”. Add up to four more images as references and name them by position (image 2, image 3…). Set magic_prompt to off for verbatim edit instructions.
Leave aspect_ratio at its default, auto. With images, the output follows the first image (a 16:9 input returns a 16:9 output). Without images, auto lets the model pick the canvas that best fits the prompt. Any other value reshapes the output to that ratio.
very_low is only available when you send images (editing). A text-only request with quality: very_low is rejected with a 400 error and is not charged. Use low or higher for text-to-image.
PNG, JPEG and WebP, as a public HTTPS URL or a base64-encoded string, with an aspect ratio between 1:6 and 6:1. Plain http:// URLs are rejected.
The price depends only on quality: very_low, low, medium or high. Input images, aspect ratio and Magic Prompt do not change it. See the Pricing page for current rates.
Rate limits depend on your subscription tier. See Rate Limits for current limits.
Ideogram 4.5 accepts input images, so it can edit and use references, and prices by quality (four tiers). Ideogram 4 is text-to-image only, prices by rendering_speed (three tiers) and also accepts a negative_prompt.

​
Best practices

  • Be explicit about what stays: In edits, say what to change and add “keep everything else unchanged”
  • Name images by position: image 1 is the one being edited; refer to references as image 2, image 3…
  • Keep auto for edits: It preserves the proportions of the input image
  • Iterate cheaply: Explore edits with very_low or low and render the final asset with high
  • Production integration: Use webhook_url instead of polling for scalable applications
  • Error handling: Implement retry logic with exponential backoff for 503 errors