Hanfu Portraits

The hanfu register

Hanfu is three thousand years of cut held in one silhouette — the cross-collar, the flowing sleeve, the line that falls rather than clings. This pack walks three dynasties. The first is Tang: a qixiong ruqun, a sheer cross-collar top over a chest-high silk skirt with a wide-sleeved pibo shawl, crimson, sapphire and celadon, a jade hairpin and a gold buyao step-shake, in a garden pavilion above a lotus pond. The second is Ming: an aoqun, a stand-collar ao jacket with small gilt buttons over a pleated qun skirt, navy, vermilion and forest-green with silver and gold embroidery, a phoenix hairpin, before a painted gold-leaf folding screen. The third is Song: a beizi, a slim long open-front jacket over a cream ruqun, pale sage, dove grey and muted indigo, restrained, in a scholar's interior with a calligraphy scroll. Three takes each, colourway and pose shifting, so the nine portraits read like a real shoot across three eras in one face.

These are photographs, not paintings — and not anime. The register leans elegant and historical, but the output is photoreal: your face and your features, the silk's quiet sheen against raised gold thread, the jade cool, the pavilion and the screen soft behind you. It borrows the cut of hanfu, the dynasty-correct grammar, and the register of a Chinese portrait shoot; it does not turn you into a cartoon, a game render, or a costume sticker. The closer your upload sits to a clear, front-on photo, the sharper the likeness comes back.

How to photograph yourself for hanfu portraits

Good input, good portrait. The pack forgives an imperfect photo — the free preview shows you the result in seconds — but four small things sharpen it.

1. Soft, even lightFace a window with indirect daylight on you. Skip the overhead bulb and hard noon sun — both flatten the face and kill the modelling the pack is built around.
2. Front or three-quarter, at eye levelHold the camera level with your eyes. No selfie up-angle, no looking down — the hanfu register wants a composed face square to the lens.
3. Plain wall behind youLet the pack build the garden pavilion, the painted screen and the scholar's room. A clear backdrop means the whole likeness budget goes to your face.
4. Hair off the faceThe hairpin and buyao sit at the hairline, set well back. Hair pulled away from your face lets the pack place the ornament and the cross-collar cleanly instead of inventing what it can't see.

Frequently asked questions

What's the difference between the free preview and the paid pack?

The preview is one hanfu portrait built from your photo, shown in seconds — a quick sketch, close but not you yet. To unlock the full pack you train your model: 12–20 photos, and we email you when your model is ready. From then on, every portrait — three looks (Tang ruqun, Ming aoqun, Song beizi), three takes each — renders from your model in about ten seconds.

Will it actually look like me?

That's the whole job. Upload one clear, front-facing photo and the pack keeps your face, your features, your skin and hair — then dresses them in the hanfu. The free preview shows you exactly how close the likeness is before you spend a cent.

Is this photoreal, or is it an anime or game-style render?

Photoreal. The register is historical and elegant, but the output is a photograph of you — real silk with its sheen, raised embroidery, jade that catches a cool highlight, real light. It is not a cartoon, an illustration, or a game render. If you want a drawn hanfu, this isn't that pack; if you want a photographic portrait in real hanfu, it is.

How is this different from Midjourney or a free AI photo app?

Midjourney will paint you a beautiful woman in hanfu — just not you; it invents a face. The free face-swap apps keep your face but render it plastic, the embroidery smeared and the collar melting. This pack does one narrow job well: your real likeness, in real hanfu, photoreal, nine ways.

Are the three looks historically distinct?

Yes — that's the point. Tang gets the qixiong ruqun and the pibo shawl, Ming the stand-collar aoqun with gilt buttons, Song the slim beizi over a cream ruqun. Three dynasties, three silhouettes, all on your one face, so you see the spread rather than one guess.

Do I need several photos, or is one enough?

One clear, front-facing photo is enough for your free preview — a couple of extra angles nudge that likeness a little closer too. To unlock the full pack, you train your model on 12–20 photos; that's what every portrait afterward renders from.

What if my photo is a casual phone selfie?

The free preview tells you in seconds whether the likeness lands. If it doesn't, the prep tips above are the fastest fix — usually it's the lighting. One good photo beats five bad ones.

What happens to the photo I upload?

Your preview photo is processed on European servers. If you train a model, those 12–20 photos are used to train your model and nothing else — delete your model and they go with it. Never sold, never shared with third parties.

When the hanfu pack fits

• A hanfu-revival portrait for the feed, in the dynasty register you actually love — Tang, Ming, or Song.
• The cultural-photography trip you didn't book, or the garden shoot the weather cancelled.
• A Lunar New Year or festival look with real historical weight rather than a generic filter.
• Diaspora users picturing themselves in the dress of their heritage, from a flat in Vancouver, London, or Singapore.
• A profile picture that signals taste and history instead of another phone selfie.
• A gift for the friend deep in the hanfu world who would never sit for the shoot themselves.

About narcis

Most AI portrait tools ship every pose, profession, era, and aesthetic in one giant catalog. The output looks like every other AI photo on the internet — a glaze over a stock face. Narcis goes the other way. One tool, one job: turn photos of you into portraits worth keeping. Drop a selfie and your free preview streams in seconds — a quick sketch: close, not you yet. For portraits that are truly you, train your own model on 12 to 20 photos; it trains in the background, and we email you when it is ready. Train once, render everywhere — this pack, every other pack, and the Studio all come out of that one model, about ten seconds a portrait, and it stays yours to retrain or delete. Each pack is hand-curated against actual references — paintings, photographers, registers — and tested on real faces before it ships. See a free preview before you pay, not after. Packs are one-time — no subscription to buy one; Studio, the unlimited composer, is optional at €12/month and every pack includes 30 days of it. No template generator. Built and run by one team, on our own infrastructure.