Upload one selfie, pick a genre and get a film poster with you as the lead, whether the movie is a rom-com, a horror story or an anime epic. A second pass adds a made-up title in clean lettering. Everything stays fictional: your face, an invented film, no real studios or stars.
Almost everyone has pictured their own name above a film title at least once. With an image editing model it takes one selfie and a short description: the model keeps your face, rebuilds the scene around you as a rom-com, a horror one-sheet or an anime epic, and a second pass prints a made-up title on it. You end up with a fake film poster you can post, print or give as a gift, and it stays fake on purpose.
A movie poster is a forgiving format. It is a single still image with a face, a mood and a title, and the viewer's imagination fills in the rest of the film. That is exactly the kind of picture image models are good at, which is why a fake poster works as a birthday gift, a joke for the group chat, a banner for a profile, a party invitation or the cover art for a short video you really did shoot.
What changed recently is that you can be the lead. Older text-to-image tools could invent a convincing actor but never you. Editing models such as nano banana pro take your selfie as a reference and keep the face recognizable while the costume, the lighting and the whole background change around it. On Picasso IA that model sits next to 502 others in the catalog, so you can compare how different engines read the same genre brief, and new accounts get free credits to try it before paying. How credits work is explained on the pricing page.
One boundary before we start: this page is about your own face, or about fictional characters. A poster of a friend needs that friend's yes, and posters built on a real celebrity's likeness are not what this is for.
The genre does most of the work. A poster is read in a glance, and the glance is decided by color, pose and typography long before anyone reads a word. Pick one genre per run and describe its habits in plain language: where the character stands, what the light is doing, what the title lettering looks like.
| Genre | Look to ask for | Pose and props | Title lettering |
|---|---|---|---|
| Romantic comedy | Soft pastel palette, warm daylight, city street | Mid-laugh, coffee cup, umbrella | Rounded, friendly script |
| Horror | Desaturated teal and black, hard side light, fog | Flashlight, doorway, half-turned face | Thin, cracked, widely spaced capitals |
| Anime epic | Painted sky, dramatic clouds, wind in hair | Sword or staff, low camera angle | Bold brush strokes |
| Sci-fi adventure | Orange and teal, lens flare, dusty planet | Helmet under the arm, distant ship | Wide metallic capitals |
| Western | Sepia dust, low sun, wide empty landscape | Hat, long coat, squinting stare | Slab serif, worn edges |
Two habits speed this up. First, describe the poster as a poster: portrait format, one central figure, space left at the top or bottom for a title. Models respond to that framing and keep the face large instead of drifting into a wide scene. Second, change one thing between runs. If you swap the genre, the costume and the pose at once, you cannot tell which change helped. The poster design page covers layout habits in more depth, and portrait styles is a good place to find vocabulary for lighting and mood.
This is the route from a selfie to a finished poster. It assumes one photo and a genre you already have in mind.
Lettering is where fake posters usually give themselves away. Image models used to produce soup instead of words; newer ones are much better, but long text still goes wrong, so the workflow is built around keeping text short and separate from the face.
GPT Image 2.5 Flare is the practical choice for the title pass. Its model page lists image editing and clean on-image text as headline features, so you can hand it your finished poster and ask it to add a title without redrawing your face. Write the exact words in quotes, say where they sit (top third, centered, bottom edge) and describe the style rather than naming a font: thin and widely spaced for horror, chunky and glossy for a comedy.
Ideogram v4 Quality is worth knowing about for typography-heavy ideas. It is a text-to-image model with a natural-language prompt and no reference-image input in its settings, so it cannot carry your selfie. Use it when the poster does not need your face, such as a fictional character, or when you want to explore title treatments first and then recreate the look in the editing pass.
A few habits keep the lettering clean. Limit the poster to a title and, at most, a short tagline. Skip the tiny billing block at the bottom: it is exactly the kind of small print models turn into gibberish, and a poster reads fine without it. If you want a credit line, use your own name in a larger size, and add it as a separate edit so a bad result is easy to undo. If one letter is wrong, regenerate the pass instead of asking for a patch on a single glyph; it is usually faster.
The fun of the format is that nobody mistakes it for a real release, and a few rules keep it that way. Invent the title. Do not use the name of an existing film, a real studio logo, a real franchise banner or the name of an actor or director. Prompts that ask for a mood are fine (a moody neo-noir thriller), but a prompt that names a specific title or star is asking the model to copy somebody else's work or likeness.
The safest poster is built from three invented things: a made-up title, a made-up tagline and a made-up studio line. If you can find any of the three in a cinema listing, change it before you post it.
The same goes for faces. Your own selfie is fair game. A friend's is fair game with their permission, and it is polite to show them the result before it goes anywhere public. Anyone else, and above all a real public figure, should not appear on a poster you make here, even as a joke. If you want a cast, generate fictional characters instead and keep them consistent with the approach on the consistent AI characters page.
A poster generator can produce real-looking results, and it helps to know where it stops. These are the five places it most often does.
None of this makes the idea less fun. It just means the last pass, looking closely at your own face and the lettering, is yours.
New Picasso IA accounts get free credits, and a poster is a modest job: one editing pass for the scene and one for the title. That is usually enough to test a genre or two. Each generation costs credits and the amount depends on the model and the settings you choose, so for plan details check the pricing page rather than trusting a number written here.
For the scene, nano banana pro is a good first choice because it takes your selfie as a reference image and holds on to the face while the setting changes. For the title, GPT Image 2.5 Flare is built to render clean on-image text. There is no single winner, so run the same brief through two models from the catalog and keep whichever result looks most like you.
No. A clear phone selfie in even daylight is enough, and often better than a stylized portrait because the model has less to second-guess. Face the camera or turn slightly, keep your eyes visible and avoid filters, sunglasses and heavy shadows across the face. If the first result does not look like you, try a different photo before you try a different prompt.
Usually because there is too much of it. Models handle a short title far better than a tagline plus credits plus a release date. Put the exact words in quotes, say where they go and describe the lettering style. If a letter is off, regenerate the title pass rather than patching it, and consider leaving out the small print altogether, since few people read it on a joke poster anyway.
A friend, only with their permission, and it is best to show them the result before it is shared. A celebrity or any real public figure, no: a poster that uses a real person's likeness can mislead people and cause harm, even when meant as a joke. If you want a cast of characters, generate fictional ones. They can still look striking and nobody has to be asked.
It is better not to. A real title, studio logo or franchise banner makes the poster look like an official piece of key art, and those marks belong to other people. Invent a title and a tagline instead, and describe the genre by its mood: a rain-soaked neo-noir, a sunny small-town romance, a sprawling anime fantasy. You keep the feel and lose the legal and honesty problems.
Yes, and for some projects it is the better route. Describe the character in detail, generate it with a text-to-image model such as Ideogram v4 Quality if you want strong typography ideas, and then reuse that image as a reference for later posters so the character stays recognizable. The same method is used for comics and game art, covered on the page about consistent AI characters.
Most one-sheet posters are tall, and a 2:3 portrait ratio is a close match that nano banana pro offers directly. Choose the highest resolution the model allows if you plan to print, and leave blank space at the top or bottom for the title. Printers have their own size and margin requirements, so check what your shop accepts before you export the final file.
Yes, as long as the file is large enough for the size you want. Generate at the highest resolution the model offers, run a super resolution model if the edges look soft, and check the title lettering at full size. Ask your print shop for its requirements on resolution and bleed, because they vary. Keep in mind that a poster with an invented title is a personal item, not official key art.
The poster is a good starting frame. Animate it as an image-to-video clip in the toolkit, or build a short sequence of stills in the same genre and cut them together with music. The pages on turning photos into videos and AI short films and trailers walk through the options, and the same rule applies: invented title, your own face.
Ready to see yourself in the lead? Open the Picasso IA toolkit, attach one selfie and run the first genre.
Choose your language
