Meta Marketing API: GenAI features vs an uploaded Sume file

Meta's Ad Creative Generative AI Terms apply when you use Meta's GenAI product features through the Self-Serve Ad Interfaces or the Marketing API. A script that renders a video with Sume and uploads the finished file is two steps, and only a call to a Meta GenAI feature is the one the page describes. Log the two steps separately in your pipeline so you can say which was which.

What does the page say about scope? (read 2026-10-09)

How should the pipeline be shaped?

Keep a render stage and an upload stage. The render stage calls Sume: trim with video_trim or assemble with Timeline 1.0, then probe the output with video_inspect. The upload stage hands the final MP4 URL to Meta. If you add a call to a Meta GenAI feature, make it a third stage with its own log line.

What should each stage record?

• Render: Sume job ID, tool name, output URL, and the probe's duration, size and codecs.
• Upload: the Meta creative ID that the file became, and the time.
• GenAI call, if any: which Meta feature, which Input file, and the output's creative ID.
• A flag for any output you handled that came from Meta's generative features, so no later step strips what the page says not to alter.

Does a re-encode matter?

The page names watermarks and content-authenticating metadata that may be generated with Output. Sume's docs do not describe that data, so treat any Meta GenAI output as untouchable: do not run it through a trim or filter. See what a trim or re-encode does to provenance marks. This is a production note, not legal advice.

Document the boundary in your repository, in a short README next to the pipeline code. A new team member should be able to see at a glance which function calls Sume, which uploads to Meta, and which, if any, calls a Meta generative feature. It makes terms updates easier to review.