GPT-Image 2.6 Flare & Sunburst are liveImages from $0.025 each
Back to templates
Text template·Structured output

GPT-6.1 Creative Brief JSON

Turn a campaign request into a strict, editor-friendly JSON brief with timed scenes, continuity, audio, and open questions.

Creative briefsStructured outputs
GPT-6.1 Creative Brief JSONText

STARTING PROMPT

Reusable prompt

The original English prompt is kept unchanged so you can reproduce the GPT-6.1 Creative Brief JSON example.

Original English prompt

Convert the creative request below into a strict JSON production brief for a small video team. Request: [paste the campaign request] Return JSON only, with this exact top-level shape: { "objective": "", "audience": [], "duration_seconds": 15, "aspect_ratio": "16:9", "scenes": [{"start": 0, "end": 0, "purpose": "", "action": "", "camera": "", "look": "", "audio": "", "continuity": ""}], "on_screen_text": [], "negative_prompt": [], "open_questions": [] } Use numbers for timecodes, keep scenes in chronological order, and make every field useful to a downstream editor or video model. If the request omits a fact, add a concise question to open_questions rather than fabricating one.

What came back when we ran it

We ran the template once on 2026-10-27 through POST /v1/chat/completions with gpt-6.1-sol at default settings, filling the request line with a 15-second launch teaser for a reusable steel water bottle aimed at trail runners, shot on a mountain trail at sunrise, with no voiceover and an end on the product. The reply took 11 seconds and 638 output tokens.

  • It returned plain JSON with exactly the 8 top-level keys and no code fence, so it parsed on the first try.
  • Three scenes ran 0 to 5, 5 to 10 and 10 to 15 seconds, matching the 15-second duration, and every scene kept "no voiceover" in its audio field.
  • open_questions held three questions instead of invented facts: the bottle design and branding, whether to use music, and whether any copy or logo should appear. on_screen_text stayed empty.
  • It kept the 16:9 aspect ratio from the template shape although the request never named one.

Make it yours

  1. 01

    Change the duration

    Replace the 15 in "duration_seconds": 15 with the length you need, for example 30. In our run the three scenes added up to exactly the 15 seconds given, so this number sets the timeline the scenes are cut to.

  2. 02

    Set a fixed aspect ratio

    Change "aspect_ratio": "16:9" to the delivery ratio, for example "9:16" for a vertical cut, so the editor does not inherit a wide frame by default. The reply kept 16:9 even though the request never asked for it.

  3. 03

    Deliver the branding facts

    Answer the likely open questions inside the request, for example "branding: matte forest green bottle, logo small on the base". The brief can then use the fact instead of listing it in open_questions, which exists to hold gaps.

Check before you rely on it

  • Effort levels none and minimal are rejected by gpt-6.1-sol, and temperature and top_p must be dropped while reasoning is on. An agent built to switch reasoning off needs a new setting before this brief will run.
  • The knowledge cutoff is April 30, 2026. A brief that relies on a release from after that date has to carry the facts in the request, since the model cannot supply them from memory.
  • Validate the JSON in code before an editor relies on it. Our one run parsed on the first try, but check the eight keys and that the scene timecodes add up to duration_seconds on every reply.

Run it through the API

Built from the template's model id, prompt and settings. Set OMNIAKEY_API_KEY in your environment first.

cURL

curl --fail-with-body https://api.omniakey.com/v1/chat/completions \ -H "Authorization: Bearer $OMNIAKEY_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-6.1-sol", "messages": [ { "role": "user", "content": "Convert the creative request below into a strict JSON production brief for a small video team.\n\nRequest: [paste the campaign request]\n\nReturn JSON only, with this exact top-level shape:\n{\n \"objective\": \"\",\n \"audience\": [],\n \"duration_seconds\": 15,\n \"aspect_ratio\": \"16:9\",\n \"scenes\": [{\"start\": 0, \"end\": 0, \"purpose\": \"\", \"action\": \"\", \"camera\": \"\", \"look\": \"\", \"audio\": \"\", \"continuity\": \"\"}],\n \"on_screen_text\": [],\n \"negative_prompt\": [],\n \"open_questions\": []\n}\n\nUse numbers for timecodes, keep scenes in chronological order, and make every field useful to a downstream editor or video model. If the request omits a fact, add a concise question to open_questions rather than fabricating one." } ] }'

Questions

Why does the brief keep 16:9 when I did not ask?

Because the schema in the prompt sets "aspect_ratio": "16:9" as a fixed value, and the run on 2026-10-27 kept it although the request named no ratio. Change that value in the template when the delivery is vertical or square.

How do I call this template from the API?

Send POST /v1/chat/completions with model gpt-6.1-sol and the template text plus your campaign request as the user message. In our run the reply came back as plain JSON with the 8 top-level keys, so it parsed directly. Billing is per input and output token, with the rates on the model page.

What happens when my request leaves details out?

The brief puts them in open_questions instead of inventing them. In our run it asked about bottle design and branding, about music, and about on-screen copy, while on_screen_text stayed empty. Answer those questions inside the request and the field list gets shorter.

ONE ACCOUNT · ONE API

Create with OmniaKey

Start with the prompt and settings above, then adapt the idea to your project.

Get an API key