> An installable agent skill for AI Content Drop. Machine-readable copy: > https://aicontentdrop.com/skills/marketing-studio-ad/SKILL.md. > Ships with the aicontentdrop package, and is listed in > /skills/index.json.

Making a finished ad with the studio tools

The studio tools turn a brief into a captioned, finished ad: list_ad_formats, list_ugc_avatars, estimate_ad_cost, write_ad_storyboard, generate_ad, get_ad_job, generate_ugc_video, generate_avatar, import_asset, list_brand_profiles and import_brand_profile. They appear on tools/list where the studio surface is enabled. If they are absent, say so and stop; do not try to assemble an ad out of generate_video calls instead.

Three of them spend credits: generate_ad, generate_ugc_video and generate_avatar. Everything below exists so the person knows the number before any of those three runs.

1. Brief first, always

An ad job is only as good as the brief, and every field of it is a question the person can answer in one line. Collect, in this order:

1. Brand — name, and a brand profile if one exists: list_brand_profiles

returns the saved ones; import_brand_profile with a website URL makes one.

2. Value props — two or three, concrete, in the customer's words.
3. Call to action — the one thing the ad asks for.
4. Format — from list_ad_formats, never from memory. Each format carries

an aspect ratio, a duration range and a shot range; pick within them.

5. Duration — inside the format's range; shorter holds attention better.

Read the brief back in four lines before going further. A wrong brief costs a whole render; a wrong line costs a sentence.

2. Assets by URL

generate_ad takes avatar_url and product_image_url, and both must live on the platform. Anything else — a Dropbox link, a CDN, the person's own site — goes through import_asset first, which copies a public https: image (JPEG, PNG or WebP, up to 10 MB) and returns the URL to pass on.

No product image? Set needs_product_gen. No avatar? Either set needs_avatar_gen or make one with generate_avatar; both add to the quote, so neither happens silently.

3. Storyboard before money

write_ad_storyboard costs nothing. Run it with the brief, show the shot list, and adjust the brief until the person is happy with it. Price and generate from the approved storyboard, not from the brief alone. Every ad quote is an upper bound (basis: "estimate_upper_bound"): the charge on success is never higher than credits_total. The storyboard quote is the tighter of the two, because a brief-only quote has to assume the most expensive shot mix for the requested quality and tier.

4. Quote, then ask once

Call estimate_ad_cost with exactly the fields you will send to generate_ad — storyboard or brief, quality, model_tier, persona_mode, the two needs_* flags. It returns credits_total, the lines behind it, a basis (flat or estimate_upper_bound) and a quote_id valid for 15 minutes.

Then the paid-approval question, in words, with the number in it:

> This ad will use up to 210 credits: Storyboard planning 1, Video shots × 3 > 204, Music bed 5. Shall I start it?

Those are the lines of a real quote for a three-shot storyboard at fast quality; premium adds a Premium lip-sync × 3 line (60) for a ceiling of

1. Read the lines from the quote you were given, never from memory.

Ask once per task. The default is to confirm each spend; only when the person has said clearly that the whole task may proceed without further questions do you skip the question for later spends in the same task. An ambiguous answer is not a yes.

Credits are charged only when the job succeeds, and the account is charged what the job actually used, so the quote is the ceiling of the surprise, not a prepayment.

5. Start, and handle the quote refusals

generate_ad with the quote_id and the same fields returns job_id, run_id, credits_quoted and poll_after_seconds. If replayed is true the quote had already started a job: report that job, do not start another.

6. Poll with patience, not persistence

Call get_ad_job with the job_id every 15 seconds — retry_after_seconds says so on every non-terminal answer. Report the stage and percent when they change, not on every poll. Stop polling after 15 minutes and hand back the job_id so the person can check later; never start a second job because the first one is slow.

succeeded brings video_url (captions burnt in), storyboard_sheet_url, duration_seconds and credits_used. failed brings error and charged nothing.

7. When a plan is needed

ENTITLEMENT_REQUIRED means the account cannot run studio jobs on its current plan. Say that, give the info_url the refusal carries (https://aicontentdrop.com/plans), and say nothing else about plans — no comparison, no recommendation, no price.

UGC avatar videos

list_ugc_avatars returns the presenter catalogue with a portrait for each; the person names one and you pass its id. A photo of a real person can be used instead, and that is where image_rights_consent comes from: it is required in the schema, it defaults to nothing, and it must be true before generate_ugc_video will run. Ask in words — "Do you have the right to use this person's likeness in an ad?" — and pass true only after a yes. Never set it on the person's behalf.

The flow is otherwise the same: estimate_ad_cost with kind: "ugc" (a flat price), the paid-approval question, generate_ugc_video, then poll get_generation with the returned id.

What to hand back

The final video_url, the storyboard sheet, the format and duration used, credits_used, and the brief as it was actually sent. A person who can see the brief can direct the next version themselves.

---

Install the plugin: https://aicontentdrop.com/plugin · All skills: https://aicontentdrop.com/skills