Your first MuseFrame piece
Most AI image tools want you to learn prompt writing first. MuseFrame skips that: you pick a direction from thirty styles across seven exhibitions, and each style already has its intent and constraints written down. Here is how to get a good first result.
1. Which originals work
Most of what the result can be is decided before you upload. Run your candidate past these:
- Sharp focus. Zoom in until you can see the eyes, or the texture of the subject. A soft original doesn't get repaired; it gets stylised and stays soft.
- A big enough subject. If a face is a few dozen pixels across, no model has anything to hold on to. Half-length beats a wide group shot by a mile.
- No extreme light. A face lost in crushed shadows, or highlights blown to white, leaves no information to work with; see how to read a histogram.
- Fewer people. Every extra person is one more face to keep recognisable. Start with a single portrait or a still life.
- Not a screenshot or a much-recompressed copy. The detail is already gone.
2. Choosing a direction, not a prompt
The thirty styles sit in seven exhibitions, each suited to certain subjects. For a first try, match the subject:
| Your photo | Start here | Roughly |
|---|---|---|
| Headshots, family portraits | Quiet Portraits | Window light, restrained colour, room to breathe |
| Still life, anything papery | Printed Matter | Misregistered ink, cut paper, letterpress bite |
| Landscape, travel | Dream Geography | Mist layers, blue distance, lifted pastel weather |
| You want a poster, a strong statement | Graphic Light | Hard light, flat colour, decisive shadow |
| Fabric, objects, pets | Material Studies | Surfaces you could touch: weave, mineral, wax, charcoal |
| Street, cinematic mood | Small Cinemas | Grey-green air, tungsten rooms, last platform |
| You want type in the frame | Small Press | Zine poster, cover story, ink and seal |
Styles marked Premium need a Creator subscription. Sign up with your email and you get 3 free pieces in any other style, so don't agonise. Pick the closest match and you'll know in two minutes.
MuseFrame turns one photograph into a piece that still looks like you: choose one of 30 styles in 7 exhibitions, about two minutes. EXIF and GPS stripped on device.
3 free when you sign up with email3. What the three choices control
After the direction there are only three settings, and people mix them up. Each one controls something different:
- Style strength sets how far the result moves from the original. Low looks like a grade; high looks like a repaint.
- Subject fidelity sets how tightly the face and identity are held. For portraits of people you know, keep it high.
- Composition is either kept or rebuilt. Kept leaves the structure of the frame alone; rebuilt lets the style move things around. If the original is well composed, keep it.
4. What to check in the result
On the result page you can hold to see the original and drag a slider to compare region by region. Check in this order:
- The face. Proportions, expression, hairline. Is it still that person?
- Hands and glasses. These go wrong most often; zoom in on both.
- Any text in the frame. Signs and book spines tend to turn into letter-like shapes that don't read.
- New things in the background. Sweep the edges against the original.
If it isn't working, change direction rather than regenerating. Rerunning the same style gives you another roll of the same taste; a different exhibition sees the picture differently.
5. Saving, sharing and printing
- The 3 free pieces are complete, with no watermark, at standard resolution.
- Prints need a bigger file. Standard resolution is fine for posting; high-resolution output for prints above A4 comes with the subscription.
- Prints come out darker than the screen. A screen emits light; paper reflects it. Make a small test print first, or lift the brightness a touch.
- The photo is re-encoded on your device, and GPS and camera-model metadata are stripped before it leaves the phone. Generation runs on a server, so unlike a purely local tool such as Mark, the picture does travel. Originals are cleared after 30 days by default, and deleting a project removes its images within 24 hours. Background in what is hidden in a photo.
The web app is at museframe.lenscript.cn/app and works in a phone browser. A piece usually takes about two minutes, longer at busy times. You can leave the page and pick up the result from your project list.
The MuseFrame site has a page each on choosing a direction, pulling the result back towards the real face, and printing: choosing a style direction, why AI portraits stop looking like you, printing: ratio, resolution and framing.
6. FAQ
Do I have to write prompts?
No. Each style is a written specification. You choose one direction from thirty styles in seven exhibitions, set style strength, subject fidelity and composition, and the specification handles the rest.
Is it still me afterwards?
That's a hard constraint in every style: facial structure, expression, pose, hair, skin tone and accessories such as glasses are kept. No beautifying, no swapping faces, no adding or removing people, no objects that weren't in the original. Hold the result at any time to check it against the original. If the original is soft or the face is tiny, though, no constraint can save it, which is why picking the photo matters.
Will my photo be used to train a model?
No. It is used to make the piece you asked for: no model training, no face database, no resale. Metadata is stripped on your device before upload, and originals are cleared automatically after 30 days.
Do the 3 free pieces you get on sign-up have a watermark or limits?
No watermark. Sign up with your email and you get 3 free pieces, each one complete and at standard resolution. The one limit is that Premium styles need a Creator subscription; all 3 can use any other style. And a generation that fails or is blocked by the safety policy doesn't cost you anything.