# How to: Portraits and avatars

> Head-and-shoulder photos of people for testimonial placeholders, team pages, personas and demo data.

## When to use it

Testimonial placeholders, a team page before the photos arrive, user personas, avatars for demo data and mockups.
Your agent can draw flat cartoon faces in code, which work as illustrated placeholders but don't read as people. In
our side-by-side test the avatars were one of the jobs vincevangen won outright, and a blind judge agreed ([how we tested](https://vincevangen.com/compare#our-test)).

Use them as placeholders and demo data. Presenting a generated face as a real customer misleads the people reading
your page, and the Final tier's model terms ask you not to pass outputs off as real photographs.

## What to ask your agent

> Three testimonial avatars for the landing page: a barista in her thirties, a retired man in his sixties, a student
> in his twenties. Head and shoulders, safe for a circle crop.

## What your agent sends

One [`vvgen_generate_asset`](https://vincevangen.com/docs/tools.md#vvgen_generate_asset) call per person, with the `avatar` preset. `n` gives
variants of the same description, so three people means three calls. The first one from our test:

```json
{
  "preset": "avatar",
  "subject": "a woman in her 30s with curly dark hair and a denim barista apron, warm smile, softly blurred café background",
  "seed": 7
}
```

The preset frames the face in the middle half of a square so a circular crop keeps the whole head, and sets soft
frontal light (`lighting` changes it). Describe age, hair, clothes and setting in the subject; vary them on purpose
across the set.

## What comes back

A 1024 × 1024 master with 256, 128 and 64 px exports beside it, a preview, the credits charged and the model's
licence. The three above were made by the open-source engine's local default model at the time (FLUX.2 [dev]); the
hosted Final tier runs FLUX.2 [pro], so your results will look different from these.

Ask your agent to check hands, ears and glasses in the preview, and to re-run with a new seed rather than describe the
fix in negatives ("no glasses" tends to add glasses).

## In Cursor and Codex

Same tools, same arguments: see the [Cursor](https://vincevangen.com/docs/quickstart/cursor.md) and [Codex](https://vincevangen.com/docs/quickstart/codex.md) quickstarts.

## Results

- A barista in her thirties. Image: https://vincevangen.com/how-to/portrait-1.webp (Photo portrait of a woman in her thirties with curly dark hair and a denim barista apron)
- A man in his sixties. Image: https://vincevangen.com/how-to/portrait-2.webp (Photo portrait of an East Asian man in his sixties with grey hair and round glasses)
- A student in his twenties. All three: avatar preset, seed 7, 12 October 2026. Image: https://vincevangen.com/how-to/portrait-3.webp (Photo portrait of a young Black man with a short beard in a mustard-yellow hoodie)

## Cost

- One portrait: 6 credits ($0.06)
- Three portraits, one call each: 18 credits ($0.18)