Skip to main content

Recraft V4.1

The latest Recraft generation. Design-oriented output with legible in-image text, two quality tiers, and a literal utility mode for diagrams and mockups.
Recraft V4.1 is a text-to-image API built for design work rather than photography: posters, signage, packaging, editorial layouts and UI mockups. Its main strength is rendering text that is actually readable and correctly spelled inside the image. Two parameters shape the output: tier selects quality and size (and sets the price), and variant selects how literally the model follows your prompt. For the previous generation, see Recraft V4 and Recraft V4 Pro.

​
Key capabilities

  • Two quality tiers: tier accepts mid and pro. pro doubles both output dimensions — four times the pixels — and costs more credits
  • Standard and utility modes: variant accepts standard and utility. standard lets the model apply its own art direction; utility renders exactly what you describe, with no invented colour or ornament. The variant does not change the price
  • In-image text: renders legible words and short phrases — signage, labels, headlines — which most diffusion models struggle with
  • Eleven aspect ratios: from vertical_1_2 to horizontal_2_1, each with a documented pixel size per tier
  • Optional prompt rewriting: enhance_prompt is on by default and expands your prompt into a richer design brief, preserving any text the image must render. Turn it off for a deterministic input
  • Async processing: every request returns a task ID immediately. Poll the task endpoint or supply webhook_url to be notified on completion

​
Use cases

  • Posters and signage: designs whose wording has to be readable and correctly spelled
  • Packaging mockups: labels and boxes with real text rather than text-shaped noise
  • UI and wireframes: with variant: "utility", clean interface mockups without invented styling
  • Diagrams and schematics: literal renders that follow the brief instead of decorating it
  • Brand and editorial assets: with variant: "standard", finished pieces with their own art direction

​
Generate images with Recraft V4.1

Submit a prompt to create a generation task. The API responds with a task ID; collect the result by polling or via webhook.

POST /v1/ai/text-to-image/recraft-v4-1

Create a new image generation task

GET /v1/ai/text-to-image/recraft-v4-1

List all Recraft V4.1 tasks with status

GET /v1/ai/text-to-image/recraft-v4-1/{task-id}

Get task status and results by ID

​
Parameters

​
Output sizes

Recraft renders from a fixed set of sizes. The values below are the exact pixel dimensions each aspect_ratio produces, and they are not always the mathematical ratio the name suggests — widescreen_16_9 renders 1344x768, which is 7:4.

​
Pricing

tier sets the price. tier: "mid" is billed as Recraft V4 and tier: "pro" as Recraft V4 Pro. variant and aspect_ratio do not change the cost. One image is generated per request. The endpoint reference pages above are generated from the OpenAPI specification and are the authoritative, complete parameter list.

​
Frequently Asked Questions

Recraft V4.1 is a text-to-image API aimed at design output rather than photography. You submit a text prompt and receive a task ID immediately. Poll the task endpoint or configure a webhook to receive the generated image when processing completes.
standard lets the model apply its own visual judgement — palette, decoration, display typography — which suits finished design pieces. utility renders exactly what you describe, with no invented colour or ornament, which suits diagrams, wireframes and UI mockups where the output has to follow the brief rather than reinterpret it. Both cost the same.
tier selects quality and size at once. pro renders at twice the linear dimensions of mid — four times the pixels — and is billed at the Recraft V4 Pro rate. mid is the default and is billed as Recraft V4.
Recraft generates from a fixed set of sizes, and the preset names are approximate. widescreen_16_9 returns 1344x768 at mid, which is 7:4 rather than 16:9. The size table above lists the exact dimensions for every value, so you can pick by pixels instead of by name.
It rewrites the prompt with an LLM before generation, expanding it with composition, layout and typography detail. Any text the image must display is preserved verbatim, including spelling and capitalisation. Set it to false when you need the model to receive exactly the string you sent.
One. Submit multiple requests to generate variations.
Rate limits depend on your subscription tier. See Rate Limits for current limits.
Cost depends on the tier you request. See the Pricing page for current rates and subscription options.

​
Best practices

  • Pick the variant first: utility for anything that must follow the brief exactly, standard for finished design pieces. It costs nothing to switch
  • Quote the text: put the exact string you want rendered in quotes inside the prompt, and keep it short
  • Draft at mid, finish at pro: the prompt does not need to change between tiers
  • Choose the ratio by pixels: use the size table rather than the preset name, which is approximate
  • Turn off enhance_prompt for reproducibility: when you are A/B testing prompts, the rewrite adds a variable you do not control
  • Production integration: use webhook_url instead of polling for scalable applications
  • Error handling: implement retry logic with exponential backoff for 503 errors
  • Recraft V4: the previous generation at the same sizes as tier: "mid"
  • Recraft V4 Pro: the previous generation at the same sizes as tier: "pro"
  • GPT Image 2: OpenAI generation, also strong at in-image text
  • Seedream 5 Pro: high-resolution text-to-image with a generate and edit split