Try on a haircut with AI

The same portrait edited three times by Flux Kontext, a short bob, long waves, a copper colour, with nothing else changing. The full workflow, no account needed.

Interactive demo

Real workflow, real results. Pan and zoom freely.

Use this workflow

The real workflow: the same portrait enters three Image nodes set to Flux Kontext, each with its own haircut prompt, and you compare.

At a glance

What the workflow produces: 3 images in 3:4
Canvas structure: 8 nodes wired by 6 connections
Models wired in: Flux Kontext
Cost of one full run: about 18 credits, roughly $0.22

Cut it, grow it, change the colour: the decision is made in front of the mirror, and the mirror only shows what you already have. This workflow moves it onto your photo. The portrait enters three nodes set to Flux Kontext, an editing model that modifies the image it is given instead of inventing one: a short bob, long waves, a copper red on the original cut.

What makes a try-on rather than a portrait of a stranger is the half of each prompt nobody reads: the list of what does not move. Face, features, skin, expression, head angle, earrings, shirt, pose, framing, background, light. The prompt starts with the hair and ends with that lock, and the demo shows it as is: on the three renders, only the hair changed.

The demo portrait is generated, so as to style nobody's face. Yours enters through the Media node. The three tries cost twenty-two cents, six credits per cut: far less than a cut you regret.

How it works

1
Import a front-facing photo
Hair clear of the face, no hat or hand in the hair, soft light. The model can only style what it sees: a fringe hidden behind a hand will be invented.
2
Adapt the lock to your photo
The list "same earrings, same white shirt" describes the demo person. Replace it with yours, glasses, beard, turtleneck, everything that must stay.
3
Describe the cut as you would to a hairdresser
Length, parting, texture, colour: "chin-length bob, side parting, straight, same brown". A cut's name alone is interpreted freely; four precise words are not.
4
Compare the three side by side
"Generate all" produces the three renders. Look at the face's oval and the nape, that is where a cut is judged. Keep the render you like, or change one prompt and re-run that node alone.

Variations and use cases

The same canvas, tuned for different needs: each variation is two or three lines of prompt away.

Before cutting short

The hardest decision, because it cannot be undone. The demo's bob answers that question; also try a pixie or a layered mid-length by duplicating the node, and look above all at the nape and the face's oval.

Before a colour

Blonde, copper, embraced grey, deep black: the colour alone without touching the cut, like the demo's third node. Skin and eyes stay, which is what lets you judge whether the colour suits your complexion.

A fringe, yes or no

One line in the prompt: "add a thick straight fringe at eyebrow level". It is the cheapest test in the world for the decision that causes the most hesitation.

Showing your hairdresser

The chosen render is a clear reference, with your face, not a model's. Send it to the salon before the appointment: the cut you asked for and the cut you get move closer.

The same try on several photos

One front-facing photo, one profile, one full-length: the same prompt cabled to each Media node, and the cut reads from every angle. Each render starts from its own photo, never from another render.

A professional portrait with the new cut

Once the render is chosen, the professional headshot workflow takes it as a reference for a studio profile photo, with the cut you do not have yet.

Common mistakes

Forgetting the list of what stays put

"Give me a bob" without the lock gives a woman with a bob, rarely you. The second half of the prompt, the one listing the face, the jewellery, the clothing and the background, is the one that counts.

Naming a cut without describing it

"Wolf cut" or "such actress's cut" is interpreted at random. Length, parting, texture, colour: four words the model cannot miss.

Starting from a photo with a hat or a hand

What is hidden is invented, and what is invented is not your hairline. Clear forehead, ears visible if possible, front-facing photo.

Chaining two edits

Colouring the bob render gives a coloured bob with the drift of both passes. Each try starts from the original photo: combine cut and colour in a single prompt.

Recommended models

Flux Kontext

Seedream 5 Pro

Topaz Precision

Frequently asked questions

Does the face really stay the same?

Yes when the lock is in the prompt, no when it is not. On the demo, the three renders keep the features, the expression, the earrings and the shirt. Always check the eyes and the mouth: that is where a model drifts first.

Can I try a colour alone, without changing the cut?

That is exactly the demo's third node: the prompt asks for copper "keeping exactly the same haircut, the same length and the same parting". Replace the colour, blonde, grey, black, and keep the rest of the sentence.

Does it work for a man, a beard, short hair?

Yes, the mechanics are the same. Add the beard to the lock if it must stay, or describe it in the change if it is what you are trying. On very short hair, a longer cut is invented by the model: that is normal, it has no material to work with.

How much does a full run cost?

Eighteen credits for the three tries, six per Flux Kontext render. The price shows on each node before you click. Re-running a single try only costs that one.

What render size?

Your photo's format, 3:4 in the demo. To send it to the salon, the render as is will do; to print it, enlarge it with the Enhance node in faithful mode.

Can I do it from my phone?

Yes, and it is the natural case: a front-facing selfie, imported from the gallery into the Media node on the mobile canvas. The creating on mobile guide explains the gestures.

Is the render faithful to what the real cut will give?

It is faithful to your face and to the description; it does not know the nature of your hair. Fine hair will not hold the volume of a render that has it. Take it as a reference, and talk it over with the hairdresser.

Are my photos used to train models?

No. What you upload stays in your workspace and is handed to no training. The question matters more here than elsewhere, since it is your face.

Try on a haircut with AI

Use this workflow

Free account, no card required. The workflow opens pre-filled in your canvas.