Style matching has been the weak point of AI image generation for as long as it has existed. Ask most models to follow a reference and you get the general mood, roughly the right palette, and a rendering style that drifts a little further with every image in the batch. Fine for one picture. A problem when you need forty that belong together.

Recraft V4 Styles, released in August 2026, goes at that specific problem. It generates images that follow the style of reference images you attach, holding texture, composition and rendering rather than just the overall look. It is available in Picsart, in both the AI Playground and Flow.

This guide covers what it does, how the two matching modes differ, what makes a good reference, and a set of prompts you can copy.

What Recraft V4 Styles actually does

The model takes between one and ten reference images and turns them into a style you can reuse. That style becomes a thing you attach to a generation, the same way you would attach an aspect ratio.

Two details make it different from the style controls you may already know.

It holds more than the vibe. Most style matching captures palette and a general aesthetic. V4 Styles carries texture, composition and rendering through as well, which is why a batch stays consistent instead of loosening as it goes.

It reads references from anywhere. Many models only work reliably with references generated inside that same model. V4 Styles does not care where an image came from. A frame from another generator, a page from a brand book, an asset from a campaign that shipped three years ago: if it exists as an image, the model can learn from it.

No training step, no fine-tuning, no waiting. Attach the references and generate.

Precise or Flexible: picking a mode

Style matching runs in one of two modes, set with the style match parameter.

Precise is the default, and it is the one for briefs that say “match this exactly.” It follows the reference’s texture, palette and rendering closely enough to hold across a full batch rather than a lucky first image. This is the mode the benchmark below was run on.

Flexible is the looser read. It matches the general feel of a reference rather than reproducing it, and it sits closer to the style experience Recraft V3 users already know, refined for V4. Reach for it when you want the mood of a reference and room to move.

A useful way to choose: if a client would notice the difference between your output and the reference, use Precise. If they would only notice that the two feel related, Flexible gives you more range.

What to use as a reference

The reference set does more to shape your result than the prompt does, so it is worth thinking about.

One reference works. You do not need a set. A single image is enough to build a style from.

Similar references sharpen the match. Feed it five images that share a treatment and the model gets a tighter, more confident read of what that treatment is. This is what you want for brand work, where the answer should be narrow.

Varied references widen the range. Feed it five images that differ and the model works across a broader space. This is what you want for exploration, where the answer should have room in it.

Ten is the ceiling. Most jobs do not need it.

The four parts of a style-matched prompt

With a style attached, the prompt stops carrying the look and starts carrying everything else. Four things are worth stating explicitly.

The subject

What is in the frame. Be concrete. “A ceramic pour-over coffee dripper” beats “coffee equipment,” because the style is already handling how it will be rendered and the prompt only needs to handle what it is.

The composition

Where it sits and how it is framed. Centred, three-quarter view, overhead, cropped at the edge. The style holds compositional character, but the prompt decides the actual shot.

The context

What surrounds the subject. Background, surface, props, setting. This is where a batch gains variety while staying visually of a piece.

What not to change

If something has to survive across every image, say so. Naming it costs you a few words and saves a regeneration.

The benchmark, and what it measures

Recraft ran a blind comparison across 159 real-world style references. In Precise mode, V4 Styles won 91.6% of its head-to-head matchups against seven leading image models, at the versions available on 14 August 2026.

The methodology is worth stating plainly, because it is unusually transparent for a vendor benchmark. Recraft designed the test. The judging was not theirs: five independent evaluators per matchup, sourced through a third-party crowdsourcing platform, each blind to which model produced which image. Full methodology is available on request.

Recraft is also clear about scope, and the caveat matters. The benchmark measures one thing: holding a style across a generation. It is not a measure of overall image quality, and the models it was tested against are strong models that are the reason style matching became a standard expectation in the first place. Read the number as “this is what it was built to win on,” not as a general ranking.

The four variants, and what each one costs

V4 Styles comes as four variants, split by output type and by tier.

  • Recraft V4 Styles (recraftv4_styles), the everyday raster model
  • Recraft V4 Styles Pro (recraftv4_styles_pro), the higher tier
  • Recraft V4 Styles Vector (recraftv4_styles_vector), for SVG output
  • Recraft V4 Styles Pro Vector (recraftv4_styles_pro_vector)

Pricing per generation, plus a one-off cost to create a style:

Model Raster image Vector image Style creation
V4 Styles $0.035 $0.05 $0.005
V4 Styles Pro $0.10 $0.12 $0.005

Creating a style costs half a cent and the style is reusable, so the cost sits in the generating rather than the setting up.

Worth knowing: every older model in the V4 family supports style-consistent generation too, at the same price as generating without a style. V4 Styles is the one built for the job, but you are not locked out of the others.

When a style match misses

The look is close but drifts across a batch. Check the mode. Flexible is designed to give you room, which reads as drift when you wanted a copy. Switch to Precise.

The match is too literal and every image looks the same. The opposite problem, and usually a reference set that is too narrow. Add references with more variety, or switch to Flexible.

The style is right but the subject is wrong. That is a prompt problem, not a style problem. The style is not carrying your subject description, so make the prompt more specific about what is in the frame.

The references are pulling in different directions. Ten mixed references give the model a wide space to work in, and a wide space produces variable results. Cut back to the three or four that best represent what you actually want.

A detail keeps changing that should not. Name it in the prompt. The style holds treatment, not specific content.

How to use Recraft V4 Styles in Picsart

There are two places to run it, and they suit different stages of a job.

AI Playground is where you work a style out. The other image models sit alongside it, so the same reference and prompt can go through several and you can judge the results before committing to one.

Flow is where you put it to work. Once a style is settled, building it into a workflow means a whole set comes out matching, instead of being generated one at a time and checked afterwards. That is the point where style consistency stops being a nice property and starts saving hours.

Full specifications are on the Recraft V4 Styles model page.

Tips for getting more out of V4 Styles

Tips for best results

Build the style once, then stop thinking about it

A style costs half a cent to create and is reusable. Build it properly at the start of a project, from references you have actually chosen rather than the first three to hand, and every generation afterwards inherits that decision.

Choose the reference set for the job, not for volume

More references is not better. Similar references narrow the match, varied references widen it. Decide which you want before you attach anything.

Start on Precise

It is the default for a reason, and it is the mode the benchmark result describes. Move to Flexible when you find you want more room, rather than starting loose and tightening.

Let the prompt do the work the style is not doing

With a style attached, describing the look is wasted prompt. Spend those words on subject, framing and context instead.

Use references that are already yours

Because the model reads references from anywhere, your existing brand assets are usable as-is. An old campaign, a style guide page, an illustration set nobody has touched in two years: all valid inputs, no conversion needed.

Name what has to stay fixed

The style holds the treatment. If a product, a colour or a layout element also has to survive across the batch, put it in the prompt explicitly.

Match the model to the output

Use the vector models when you need SVG, not when you want something that looks flat. Rendering style is the style’s job; file format is the model’s.

Frequently asked questions

An AI image model built for style-consistent generation. It creates a reusable style from one to ten reference images and then holds that style across every generation, preserving texture, composition and rendering rather than just the general look.

Start with one reference

The fastest way to understand what V4 Styles does is to give it something you already have. Pull an image from a campaign that shipped, attach it, and generate something new against it. If the texture and rendering come through and not just the colours, that is the difference the model was built for.

Try Recraft V4 Styles in Picsart’s AI Playground, or build it into a workflow in Flow.