Toy product video: use real demo footage, not AI play, per CPSC rules
Selling toys online: why to film the real toy, how the CPSC choking-hazard warning reaches ads, and how to burn that line onto clips with Sume cues.

A toy listing video makes a promise about a physical object. If a generator invents the play, the clip shows the toy doing something your product never did, in the hands of a child you never filmed. For toys, film the real product and use software for the assembly and the required text. This is not legal advice; check your own compliance file.
What the CPSC pages actually say
The CPSC Toy Safety business guidance says products designed or intended primarily for children 12 and under need third-party testing at a CPSC-accepted lab and a Children's Product Certificate; for toys the CPC cites 16 C.F.R. part 1250 plus the applicable ASTM F963 sections.
The page on the small parts ban and choking hazard labeling states: "Any advertising for children's toys subject to the choking hazard warning labeling requirement must use the same warning label per 16 C.F.R. § 1500.20." So the warning belongs in the ad, not only on the box. Neither page mentions AI video; the point here is that a certified toy is tested as built, and the footage should show that toy.
A split that keeps the claims honest
Shoot the demo clips on a phone with the real toy: unboxing, the one trick it does, a size shot next to a hand. Then let Sume do the parts that carry no product claim.
| Piece | Source | Sume surface | Cost |
|---|---|---|---|
| Demo clips | Your camera, real toy | None | Your time |
| Assembly of 3 to 6 clips | Real clips, imported | Timeline 1.0 | $0.10 per output minute |
| Warning line on screen | Your compliance text | Video captions with cues | $0.20 per clip up to 60 s |
| Opening packshot motion | Your product photo | Image-to-video | Per second, by model |
Burn the warning as authored cues
Caption cues skip speech-to-text and burn exactly the text you send, each with text, start and end in seconds. A cue holds up to 400 characters, a job takes up to 200 cues, and the source can be up to 60 seconds. Read the wording from your own file rather than retyping it; the script below refuses to run without it.
import os
import requests
warning = os.environ["TOY_WARNING_TEXT"]
if not warning.strip():
raise SystemExit("TOY_WARNING_TEXT is empty")
resp = requests.post(
"https://api.sume.com/v1/video-captions",
headers={
"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
"Idempotency-Key": "toy-listing-warning-001",
},
json={
"video_url": "https://media.sume.com/artifacts/example/toy.mp4",
"cues": [{"text": warning, "start": 0, "end": 20}],
},
timeout=60,
)
print(resp.status_code)
print(resp.json())Cost of one 30-second listing video
Timeline assembly of a 30-second cut reserves one output minute, $0.10 at the public rate in the Timeline docs. The warning burn is $0.20 under the captions docs. Together that is $0.30 per finished clip, with no generation spend. Confirm live rates in GET /v1/catalog, as both pages advise.
Before you publish
The result is a listing video whose footage matches the certified product and whose required line travels with it into every placement.
- Confirm the age grade and the warning wording with your lab report and CPC, not with a blog post.
- Check the warning is readable at phone size for the full clip, or at least while the toy is shown.
- If you use generated motion anywhere, keep it to the packshot, never to children playing.
- Keep the cue text in one file so every clip carries the same line.
Sources
Related posts
More in Use cases
- Training video policy change: re-render one 15 s module
A policy change in a 45-second avatar training video costs $8.28 to re-render on Sume's standard tier. Three 15-second modules cut that to $2.76.
- Transcribe a podcast back catalogue: 100 episodes cost $31
At Sume list rates a 30-minute episode costs $0.31 to transcribe: $0.30 of STT plus one split job. Cost for 4, 10, 52 and 100 episodes and a safe batch order.
- Transcribe a 2-hour recording: twelve 10-minute STT jobs, $1.24
A two-hour recording is twelve 10-minute Sume STT jobs at $0.01 a minute, cut with four Timeline audio split jobs: $1.24 at list rates, with the seam caveat.
- Translate caption cues but keep the timings: a reading-speed check
Keep each cue's start and end, swap only the text, and flag lines that become too dense. A short Python check catches it before you pay for the render.
Written by Sume