Halloween safety video with an AI avatar for parents and schools
A 21-second three-scene avatar video built from the FDA Halloween tips: costumes, candy and contact lenses. The request body, the cost by tier, and the label.
Halloween is on Saturday, 31 October 2026, which leaves a few weeks to get a safety message into parents' hands. A short avatar video works for a school newsletter or a clinic page if the script sticks to what an authority already says. The FDA's Halloween safety tips, page dated 10/19/2023 and read 2026-10-06, give three topics that fit one 21-second video with three scenes.
What the FDA page says, by scene
| Scene | FDA tip, close to its wording | Script (words) | Seconds |
|---|---|---|---|
| costumes | Wear costumes labelled flame resistant; add bright, reflective tape; wear makeup and hats rather than masks | Pick a costume with a flame resistant label, add reflective tape, and choose makeup or a hat over a mask that blocks your view. (24) | 9 |
| candy | Do not eat candy until it has been inspected at home; accept only commercially wrapped treats | Wait until you are home to eat any candy. Check every treat, and skip anything that is not commercially wrapped. (20) | 8 |
| lenses | Avoid decorative lenses that a doctor has not prescribed | Skip decorative colored contact lenses unless an eye doctor prescribed them. (11) | 4 |
The request
Each scene is a voice of type: text, and each carries a background prompt. This example repeats the same porch prompt in all three scenes so the look stays consistent. Sume checks that the estimated total sits between 4 and 60 seconds.
curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: halloween-safety-001" \
-d '{
"avatar_handle": "school_host",
"aspect_ratio": "9:16",
"quality": "standard",
"captions": { "enabled": true, "style": "slam", "language": "auto" },
"video_inputs": [
{ "id": "costumes", "voice": { "type": "text", "script": "Pick a costume with a flame resistant label, add reflective tape, and choose makeup or a hat over a mask that blocks your view." }, "background": { "type": "prompt", "prompt": "Front porch at dusk, warm light" } },
{ "id": "candy", "voice": { "type": "text", "script": "Wait until you are home to eat any candy. Check every treat, and skip anything that is not commercially wrapped." }, "background": { "type": "prompt", "prompt": "Front porch at dusk, warm light" } },
{ "id": "lenses", "voice": { "type": "text", "script": "Skip decorative colored contact lenses unless an eye doctor prescribed them." }, "background": { "type": "prompt", "prompt": "Front porch at dusk, warm light" } }
]
}'Cost and the label
At about 21 seconds without a product image, Sume's per-second Avatar Video rates give roughly $3.86 at standard ($0.184), $5.15 at plus ($0.245) and $11.55 at max ($0.55). Run a preview first if the porch look matters, then render.
Say in the video description that the presenter is AI-generated, and link the FDA page so a parent can read the source. A synthetic host is fine for reading public health tips; it should not sound like a named doctor or school official unless that person has agreed.
A parent reading a newsletter will not watch an ad. Keep the hook in the first scene concrete, such as the costume check, and put the date and school contact in the page text around the video rather than in the script. Captions help because many parents watch with the sound off, and the slam style burns them into the MP4 after generation.
The FDA page is dated 2023. Before publishing, reopen it and make sure the wording you read still matches, since agency pages change. If a tip no longer appears, drop that scene and re-run the preview, because a changed script needs a new preview.
Sources
Related posts
More in Use cases
- Homebrew beer label art with an API: the name and ABV added in code
Generate square label art on Sume with no text, then draw the beer name, style and ABV in code so every batch gets exact numbers on the same artwork.
- How to split a training lesson into avatar clips under 60 seconds
Sume refuses avatar videos over 60 seconds. Split a 6-minute lesson by words, render each part, and stitch with Timeline 1.0. Word counts and cost by tier.
- HR onboarding welcome video with an AI avatar: day-one clip
A 30-second welcome clip that lists the three things a new hire does on day one. Script size, cost per tier, and what must stay in text, not video.
- Insert a sponsor read into a podcast: TTS plus one audio concat
Generate the ad read with Sume TTS, then join it into the episode at the second you choose with one Timeline audio concat: sample-exact, $0.01 per join.
Written by Sume