TikTok AI disclosure: toggle or on-screen note for a Sume clip

TikTok Shop's AI-generated content policy accepts two ways to disclose: switch on the "AI-generated content" toggle in the posting settings, or put text, a watermark or a note inside the video. A clip made with Sume can use either. The toggle is a TikTok setting that you flip when you post; Sume does not set it. The on-screen note is part of the file, so Sume can help with that one.

The policy page prints its effective date as "09/02/2026". It does not say whether that is month-first or day-first, so this post quotes it as written.

The two accepted methods

When the policy asks you to disclose

The page asks for disclosure when the video or audio has any of these:

• synthetic faces, synthetic voices or digital humans;
• content that is fully AI-generated or significantly AI-edited;
• content realistic enough to be mistaken for real footage.

Burn the note with caption cues

Sume's video captions API takes a public HTTPS video URL. If you send cues (each with text, start and end in seconds), it burns your authored text at those times and does not run speech recognition. That is the documented path for authored overlay text, and it also works on a silent clip. Latin text with no style resolves to slam.

A standalone caption job reserves and captures $0.20 for videos of up to 60 seconds. The docs say the live price is in GET /v1/catalog.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ai-note-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/clean.mp4",
    "cues": [{ "text": "AI-generated video", "start": 0, "end": 4 }]
  }'

Or overlay a still

Timeline compose puts one still and one video on screen together. With operation: "overlay", layout.position can be top, center or bottom, and width_ratio runs 0.05 to 1 (default 0.9). The still must probe as a still, both files must be on media.sume.com (import them first), and Idempotency-Key is required. The public rate is $0.02 flat per job, and the default output is 1080x1920.

A posting-day routine

Make disclosure a fixed step so no file depends on memory. Before you post, open the clip and ask whether it has a synthetic face, a synthetic voice or a digital human, whether it is fully AI-generated or significantly AI-edited, and whether a viewer could mistake it for real footage. One yes means disclose.

Then pick one method per file and write it down: the toggle, or the burned-in note. Using both is not required by the page, but a burned-in note travels with the file if you reuse it elsewhere, while the toggle lives only on the TikTok post. If you re-cut the clip for another platform, the note survives the re-cut only if it is inside the pixels.

What stays on your side

The toggle, the decision about whether a clip needs a label, and the account consequences are yours. The page lists removal or restricted visibility, warnings, permanent bans and a Creator Health Rating deduction as enforcement. The Sume docs do not describe writing platform labels or provenance metadata, so do not assume a Sume file carries one.