TikTok split test variables: which ones need a new AI video?
TikTok lets a split test change one variable. Only Creative and Creative Assets need new media; the rest reuse one clip. Here is what to render, with Sume.
Of the eight variables TikTok lists for a split test, two need new video: Creative (ad formats, videos, video views, call-to-action copy, description copy) and Creative Assets (images, videos, ad formats, and the initial hooks). Targeting, placement, budget strategy, catalog, and bidding all reuse the same clip. TikTok also says you can only select one variable per test.
The variable list is from Split Testing Variables, read 2026-10-03.
Variable by variable
The right-hand column is advice, not a TikTok rule: it shows whether you generate anything new for that test.
| TikTok variable | TikTok's wording | New video needed? |
|---|---|---|
| Targeting | Broad vs narrow, custom audience, demographics, interests, device, location | No, one clip |
| Placement | Auto-selected and manual placements | No, one clip |
| Creative | Ad formats, videos, Video Views, CTA copy, description copy | Yes, if videos differ |
| Budget Strategy | CBO and Ad Group Level Budget | No, one clip |
| Creative Assets | Images, videos, ad formats, and the initial hooks (opening 2-3 seconds) | Yes |
| Catalog | Catalog Ads and standard non-Catalog Ads | Only the product clips |
| Bidding and Optimization | Bidding strategies, optimization goals, bid amount | No, one clip |
| Custom | Advanced measurement and cross-campaign testing | Depends on the design |
What to generate for the two creative variables
For a hook test, render the same avatar video with a different first scene each time (see Generate avatar video for video_inputs). For a whole-video test, change the script or the scene instead, and keep the opening identical so the variable stays single.
For product shots without a presenter, POST /v1/videos accepts frame_images for image-to-video; see Video generation for the fields each model supports.
Do not stack changes
If the new video also changes the call-to-action copy, you have two variables. TikTok's help text says one; keep the copy fixed and let the video carry the difference. The schedule rule is separate: TikTok says to set at least 7 days.
Sources
Related posts
More in Use cases
- Timeline fit modes: cover, contain, stretch, blur for ad clips
Sume Timeline video[].fit has four modes: cover (default), contain, stretch and blur. What each does to a clip that does not match the output frame.
- Estimate an ad video render cost before you pay: timeline plan
Sume's timeline plan endpoint is unbilled and returns billable minutes and an estimated cost before you render. Render is $0.10 per started output minute.
- Trim a UGC ad down to its hook: frame-accurate cut for $0.02
Sume video trim cuts a start/end range from a hosted clip into a new MP4 for $0.02 a job. Exact vs keyframe precision, the 0.2-900 s duration, and import first.
- UGC ad casting test: one script across three AI avatars
Test which AI spokesperson sells your script: create three Sume previews with the same copy, pick a still, then render only the winner. Curl loop and costs.
Written by Sume