Profile a brand's photography

Use a photography profile to tell your agents what a brand’s photos look like, so the images they create or pick match the brand’s photography and not just its colors and type. Each photo is read against a fixed schema of 45 fields (light, framing, subject, color, finish, and more) and its pixels are measured; the photos are then merged into one profile. A photography profile is a submission like a website extraction or a brand kit, but it takes photos instead of a URL or a PDF, and its result has kind: "photography" and its own shape.

Build a profile

Send 1 to 12 photos, each as a public https:// url or as image_b64 (PNG, JPEG, or WebP):

curl -X POST https://api.tastelabs.com/design/submissions/photography \
  -H "X-API-Key: $TASTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "photos": [
      {"url": "https://images.example.com/campaign-1.jpg"},
      {"url": "https://images.example.com/campaign-2.jpg"},
      {"url": "https://images.example.com/campaign-3.jpg"}
    ]
  }'

The API returns 202 with a submission_id. Poll GET /design/submissions/{id} until status is completed or failed, then read the profile from GET /design/submissions/{id}/result. A profile usually takes under a minute. Photography profiles appear in GET /design/submissions next to other submissions; filter them with ?kind=photography. A profile’s source_url is the first photo’s URL, or upload:// followed by the photo count when the photos were uploaded.

Each submission costs 2 credits, refunded when no profile comes back. Every request is a fresh profile: profiles are never served from cache.

How many photos to send

One photo: the profile is that photo’s own reading.
Two or more: rules keeps the values most of the photos share; with two photos, that is what both share. The rest goes to varies.

Send 3 to 8 photos that represent the brand for a profile you can generalize from.

Use cases

From a brand kit: give your agent a brand kit, have it take the brand’s photos from the kit, and send them to the photography endpoint. The profile it gets back describes how the brand’s photos look, so the new photography your agent creates stays aligned with the brand.
From a website: do the same with a website extraction. Send the site’s images and use the profile to create, for example, a new subpage whose photography matches the rest of the site. Any graphics or illustrations among those images are left out on their own.

What is left out

Every photo is read, but only photographs are merged. A photo that reads as a graphic or illustration, a photoreal render, or a composite is left out of the profile, and its outcome says why in excluded. Photos that cannot be used, such as an unreachable URL or a file that is not an image, come back as invalid. When no photo is left to merge, the submission fails and the credits are refunded. Its error tells you what to do:

`error` starts with	Meaning
`NOT_ENOUGH_PHOTOS`	None of the photos is a usable photograph. Send other photos.
`PHOTO_READS_FAILED`	The photos could not be read. Retry the same request.

The result endpoint returns 409 NOT_READY until the profile is completed, including for a failed submission. Read the outcome from GET /design/submissions/{id}.

Read the result

The following example comes from six Ai2 photos. Most rules, variations, and outcomes are omitted.

{
  "kind": "photography",
  "id": "5fb12de2-b640-4333-954a-a41703406b9a",
  "status": "completed",
  "photos": 6,
  "profile": {
    "photos": 4,
    "rules": [
      { "field": "focal_structure", "name": "Focal structure", "value": "single_dominant", "share": 0.75 },
      { "field": "vignette", "name": "Vignette", "value": "none", "share": 1.0 }
    ],
    "varies": [
      { "field": "shot_scale", "name": "Shot scale and crop margin", "values": { "close": 2, "detail": 2 } }
    ],
    "tone": [
      { "field": "exposure_key", "name": "Exposure key", "value": "mid_key", "measure": 47.29 }
    ],
    "signature": { "darkest_lightness": 1.6, "brightest_lightness": 95.0, "colourfulness": 72.0, "lightness_spread": 24.9 },
    "palette": ["#1f1914", "#504745", "#e3e2e1", "#dc9912", "#e9b73f"]
  },
  "outcomes": [
    { "index": 1, "status": "read", "main_subject": "A single yellow sunflower head with a dark central disc, seen face-on", "excluded": null },
    { "index": 3, "status": "read", "main_subject": "Undulating glossy wave surface filling the frame; no discrete object", "excluded": "photoreal_render" }
  ]
}

Field	Meaning
`photos`	Photos sent. `profile.photos` counts the ones merged.
`profile.rules`	Values most of the merged photos share, each with its `share`.
`profile.varies`	Fields where no value is shared enough to be a rule, with the number of photos per value.
`profile.tone`	Exposure, highlights, contrast, and other tone values, measured from the pixels and never read by a model. `measure` is the median the value was cut from.
`profile.signature`	How the photos render, as medians: black and white points, clipped highlights, warmth of mid-tones and shadows, colourfulness, and contrast.
`profile.palette`	Dominant colors of the photos. These are not brand colors.
`outcomes`	One entry per photo sent, in order: whether it was `read`, `invalid`, or `failed`, what it shows, and why it was `excluded`.

Give it to your agents

Hand your agent profile.rules and profile.tone as the brief for new imagery, and profile.varies as what it is free to change. Ask it to check its output against the rules, the same way it checks copy against a brand kit’s rules.

Getting Started

Use cases

Brand extractor

Brand search

Brand adherence verifier

Integrate Agents

Reference

Profile a brand's photography

Build a profile

How many photos to send

Use cases

What is left out

Read the result

Give it to your agents

​Build a profile

​How many photos to send

​Use cases

​What is left out

​Read the result

​Give it to your agents

Build a profile

How many photos to send

Use cases

What is left out

Read the result

Give it to your agents