Sume auto never names the model: what to log for TikTok AI disclosure

With sume/auto the response does not name the model family. TikTok asks if content is AI-generated, not which model made it. Log job id and date.

4 min readSume
All posts

When you send model: "sume/auto", Sume picks the model and its docs say the family is never disclosed in the response. TikTok's disclosure rules ask whether the content is AI-generated, not which model made it, so your record needs the job id, the date, and whether the toggle was on.

If you need to know the model for your own reasons, name it in the request, for example wan-3.0, instead of using auto.

What TikTok asks for

The US Shop page asks creators to disclose content fully AI-generated or significantly edited by AI, through the toggle or an on-screen note. The November 2025 newsroom post says labels can show how they were applied: AI detection, creator labels, or TikTok's own tools. Neither asks for a model name.

What to log per video (read 2026-10-07)
FieldWhyWhere from
Sume job idFinds the render laterCreate response
Requested modelRecords sume/auto or the named modelYour request
Post dateMatches the TikTok recordYour calendar
Toggle stateShows you disclosedTikTok upload screen
On-screen noteSecond route for disclosureYour edit

Log it in the request

Video generation is asynchronous: POST /v1/videos returns an id, and you poll GET /v1/videos/{id}. Send an Idempotency-Key so a retry does not create a second job, and store the id with your own tag. The jobs docs describe the job states.

curl -s https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Idempotency-Key: ep12-hook-a" \
  -H "Content-Type: application/json" \
  -d '{"model":"sume/auto","prompt":"Unboxing a ceramic mug, 9:16","duration":5}'

Pin the model when it matters

Auto is useful when you want the router to choose. Pin a model when a brief names one, when you compare cost across runs, or when you want a repeatable setup. Because no v1 model accepts a seed, pinning does not make the output repeatable, but it does make the cost and limits predictable.

If a partner asks which model

Some brands and agencies ask which AI tool made a video. With auto, Sume does not tell you the family, so answer with what you can prove: the job id, the date, and that Sume generated it. If the answer matters to the contract, pin a model by name in the request so your record is explicit.

Keep the log in plain text or a sheet with one row per post. It takes a few seconds per video and gives you a defensible answer if a label or a review question arrives later.

A small habit helps later: put the same tag in your Idempotency-Key and your spreadsheet, such as the episode number and variant letter. When a post needs checking a month on, you search for the tag and find the job.

Do not store API keys or customer data in the log; the ids and dates are enough for this purpose.

A caution on wording: do not claim in public that a clip came from a particular model unless you pinned that model. With auto, the honest statement is that the clip was generated with AI through Sume. That statement is also all TikTok's disclosure toggle needs from you, so there is no reason to guess a model name.

Sources

Related posts

More in Developers

All Developers posts

Written by Sume