AI ad disclosure record: the Sume job fields to save per clip

A per-clip record that answers a platform or client query: model, prompt, job id and date from a Sume job, plus the label decision, in one sheet row.

4 min readSume
All posts

Save one row per clip with the job id, the model, the prompt, the date, the output file, and your label decision. When a platform asks how a clip was made, you answer from the row instead of searching old chats. Sume's job records already carry the request and, for Music Router jobs, the engine that ran, so most of the sheet is copy and paste.

Platform rules are not stated here. A third-party summary on Audit Socials reports that TikTok requires a label for synthetic faces, AI voice and photorealistic AI backgrounds; read the platform's own page for the rule.

Which fields matter?

The row is short on purpose. If it is hard to fill in, it will not get filled in.

Per-clip record. Fields from Sume job pages; label column is yours. Read 2026-10-02.
FieldWhere it comes from
Job idSume job
Model idrequest.model; for Music Router jobs also request.routed_model
Prompt and inputsRequest body
Date and timeJob record
Output fileYour storage copy
Label applied?Your decision and where

Why log the routed engine for music?

For music, job.request.routed_model names the engine that ran when you use sume/music-auto, so read it once at completion and store it. For video, sume/auto is different: the docs say Sume does not disclose which family served the request, so pin a model id when you need that answer in your record.

How long should I keep it?

As long as the ad could be questioned, which is at least the life of the campaign and any contract term. Store the sheet next to the master files, not in a chat.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume