Soku AIDescribe the person and the context
Age range, wardrobe, expression, lighting and setting. "Late-forties operations manager in a warehouse, hi-vis over a polo, natural light" produces something usable; "a person" produces a stock photo.
All tools
AI Portrait Generator A Face for the Layout, Without Borrowing Someone’s
Half the ad layouts a performance team ships need a person in them — a quote card, a testimonial rail, a "meet the team" panel, a founder shot behind a headline. Sourcing that face is where the work stalls: stock looks like stock, a real customer needs a signed release, and a photoshoot needs a calendar. This composes a portrait from a written description instead, so the slot gets filled in the same session the layout was designed.Half the ad layouts a performance team ships need a person in them — a quote card, a testimonial rail, a "meet the team" panel, a founder shot behind a headline.
PortraitsSocial Proof
Three deliberate steps
Choose the crop the layout needs
Square for a quote-card avatar, portrait for a full-bleed panel, landscape when the copy sits beside the face. Composing to the final ratio beats generating a square and cropping into it afterwards.
Generate, then look at the hands and the edges
Portraits fail at the periphery long before they fail at the face. Check hands, ears, jewellery, lanyards and any text on clothing, and regenerate rather than retouch if something is wrong.
Use cases
Who reaches for this, and when
Fill testimonial and review layouts without a consent problem
Social-proof creative needs a face beside every quote, and using a real customer’s photograph is a permission question most teams answer too late — often after the ad is already live. A generated portrait fills the slot without attaching a real person’s likeness to words they did not pose for.
Prototype a campaign before anyone is cast
Concepts get killed in review because the placeholder looked cheap, not because the idea was wrong. Generating the intended demographic and wardrobe lets a layout be judged on the layout, and it costs nothing if the concept does not survive the meeting.
Localise a creative for a market you have not shot in
A campaign that tested well in one region usually needs a plausibly local face before it runs in another. Regenerating the portrait per market is dramatically faster than commissioning a regional shoot for a variant that may not earn one.
Questions before you run it
- Can it make a portrait of a specific real person?
- No, and you should not try. This composes a face from a written description rather than reproducing a likeness, so it cannot render a named individual, and generating a recognisable real person for advertising raises publicity-rights and platform-policy problems well beyond whether the image looks right. If you need a specific person in the ad, photograph them and get a release.
- Can I upload my own photo and restyle it?
- Not on this page. This is text-to-image: it starts from your description and a blank canvas, with no source photograph involved. If your goal is transforming a picture you already own — changing the background, the wardrobe or the lighting on a real headshot — that is an image-editing job and needs a tool that accepts a file as input.
- How is this different from the AI human generator?
- Framing, mainly. The human generator is aimed at a full figure inside a scene — a presenter in an environment, a lifestyle shot with a product in frame. This is aimed at the face: headshots, shoulders-up portraits and the crops that quote cards, team panels and profile slots actually consume. Both run the same underlying image model, so the difference is in how the page prompts you rather than what it can technically render.
- Do I have to disclose that the person is AI-generated?
- That depends on your jurisdiction, your platform and how the image is used, and it is genuinely a question worth asking rather than assuming. A synthetic face illustrating a generic concept is treated very differently from a synthetic face presented as a named customer giving a testimonial — the second is a claim about a real endorsement and is the one that gets accounts into trouble. Check the ad policy for the platform you are running on.
- Will I get the same face if I run it twice?
- No. Each run is independent, so the same prompt returns a different person rather than the same one again. If a campaign needs one consistent face across several placements, generate until you get the right portrait and then reuse that single file everywhere, exactly as you would with a photograph from a shoot.
Keep building
