Turn Prompts Into Art-Directed Images With Recraft V4
Last verified: November 3, 2026
Recraft V4 is Recraft's design-led image model. Recraft built it with professional designers and tuned it to make deliberate choices about composition, color relationships and detail hierarchy, not just match the words of a prompt . That means the images it returns are ready to go into a layout, a campaign or a pitch. Recraft released it in February 2026 as a ground-up rebuild of its image model. The family has four text-to-image versions: the standard V4, V4 Pro, V4 Vector and V4 Vector Pro . This page runs the standard V4.
Lettering gets the same design treatment. V4 treats typography as a structural part of the composition, not a decorative overlay . That matters for posters, packaging and covers, where words and image have to read as one piece. Recraft presents V4 as the successor to V3 for most generation tasks and keeps its highest native resolution, 2048x2048, for the V4 Pro version .
Explore Recraft AI's Models
Key Specs Before You Prompt
The controls this generator gives you and the model behaviors that shape every image.
Generation mode
Text to image
Prompt length
Up to 2,048 characters
Aspect Ratio options
1:1 (default), 3:4, 4:3, 9:16, 16:9
Images per run
1
Image type
Raster
In-image text
Short and mid-length phrases, with high fidelity to layout and typography
Check These Before You Generate
Six checks that keep lettering legible, layouts intentional and palettes cohesive.
Quote every word that should appear
Put headlines, labels and sign copy inside quotation marks. Recraft's prompting guidance links quoted wording to precise text rendering within the layout, so anything you want spelled exactly belongs in quotes.
Break long copy into headline-sized lines
A paragraph of body text goes beyond the phrase lengths the model renders most faithfully. Split it into a headline, a subline and a small label so each piece stays legible.
Say where each line of text sits
Give every text element a position: top, centered, along an edge, curved or cropped at the frame. Clear placement cues let Recraft V4 build a structured layout instead of guessing where words go.
Match Aspect Ratio to the layout you describe
The frame starts at 1:1. If your prompt asks for vertical side text, a tall poster or a widescreen scene, switch to 3:4, 9:16, 4:3 or 16:9 first, so the composition has the space your instructions assume.
Give each color a job
Assign colors to roles, such as background, type and accent, instead of just listing them. Recraft V4 is tuned around color relationships, and a defined palette gives it something coherent to balance.
Read Enhance Prompt suggestions before picking one
Suggestions turn a rough idea into a fuller brief. Before you accept one, check that it keeps your exact quoted wording and the text placement you need. Your prompt only changes when you choose a suggestion.
Recraft V4 or Flux 2: Pick the Right Generator for the Brief
This comparison puts the Recraft V4 text-to-image generator next to the Flux 2 versions you can run beside it. The aspect ratio, prompt length and images-per-run rows show the settings each generator offers here. The other rows show what Recraft and Black Forest Labs publish about their own models.
| Feature/Spec | Recraft V4 | Flux 2 |
|---|---|---|
| Creator and launch | Recraft, released February 2026 | Black Forest Labs, announced November 25, 2025 |
| Versions in the family | V4, V4 Pro, V4 Vector and V4 Vector Pro | FLUX.2 [klein] 4B and 9B, [max], [pro], [flex] and [dev] |
| Native output size | 1024x1024 (1:1), 1216x896 (4:3), 896x1216 (3:4), 1344x768 (16:9) | FLUX.2 models: up to 4 megapixels, e.g. 2048x2048 |
| Prompt following | Follows detailed direction on reflections, materials and background interactions | Follows structured prompts, with substantially better prompt adherence than FLUX.1 |
| Text in the image | Builds lettering into the layout instead of placing elements at random | Complex typography, infographics and UI mockups |
| Aspect Ratio options | 1:1 (default), 3:4, 4:3, 9:16, 16:9 | Flux 2 Pro and Flux 2 Flex: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9; Flux 2 Max: the same set plus 21:9 and 9:21 |
| Prompt length | Up to 2,048 characters | Flux 2 Pro and Flux 2 Flex: up to 2,048 characters; Flux 2 Max: up to 5,000 characters |
| Images per run | 1 | 1 on Flux 2 Pro, Flux 2 Flex and Flux 2 Max |
Put the Differences to Work
Getting words into the picture
Both creators say their models handle text, but they describe the job differently. Recraft treats lettering as part of the layout: the model works out where a headline sits relative to the subject and the empty space around it. Black Forest Labs describes FLUX.2 handling complex typography, infographics and UI mockups. For Recraft V4, the practical takeaway is copy length. Its documented strength is headline-, subline- and label-length wording, so a poster, a package front or a shop sign plays to what it is built to do.
Frame shapes and prompt room
The core frame set is the same on both generators: square, two portrait shapes and two landscape shapes. Flux 2 Pro and Flux 2 Flex also match Recraft V4's prompt box character for character. The differences are at the edges. The Flux 2 versions add 2:3 and 3:2, and Flux 2 Max adds ultra-wide and ultra-tall frames plus more than twice the prompt room. If your layout fits one of the five shared ratios and your brief fits in 2,048 characters, the frame and prompt settings give neither generator an edge.
Match the Generator to Your Frame and Brief
How to choose: Recraft V4 and Flux 2 share the basics here. Both make one image per run and have an Enhance Prompt button. Recraft V4, Flux 2 Pro and Flux 2 Flex each take a 2,048-character prompt, and all of them cover 1:1, 3:4, 4:3, 9:16 and 16:9. Choose Flux 2 Pro or Flux 2 Flex when a layout needs 2:3 or 3:2. Choose Flux 2 Max when it needs 21:9 or 9:21, or when your brief runs past 2,048 characters (up to 5,000). For the five shared frames and briefs within 2,048 characters, the two generators offer the same settings.
Go From Idea to Finished Image in Four Steps
Four steps: write, refine, frame and generate.