TikTok Shop LIVE: stills may cover only half the screen
TikTok Shop seller guidance caps stills and slideshows at 50% of the screen during a LIVE. Plan overlay layouts with motion clips from Sume instead of stills.

TikTok Shop seller guidance says that during a LIVE, still images and slideshows must not cover more than 50% of the screen. If you use AI-made product visuals as an overlay, that means a layout where half or more of the frame is moving video, or a stills panel kept under half. Sume produces clips and images, not live streams, so this is a content-planning question for the overlay assets you bring to your own streaming setup.
What the seller page says
The TikTok Shop Seller University page, read on 2026-10-03, covers video and LIVE quality. The points relevant here:
| Context | Guidance |
|---|---|
| Shoppable video | Aim for at least 3 seconds of dynamic footage; no still images or looping visuals |
| LIVE | Stills and slideshows must not cover more than 50% of the screen |
| Both | Face visible and the product shown from multiple angles expected |
Why a Timeline slideshow does not help
It is tempting to build a loop of product stills with Timeline and use it as a LIVE overlay. The Timeline docs say stills in video[] are static holds, and motion on a still is accepted and ignored with a motion_ignored warning. A render made of stills is still a slideshow in TikTok's terms, however smoothly it transitions, so it counts toward the 50% limit.
The fix is motion in the source, not in the assembly. Generate short clips for each angle so every slot holds footage.
Make the overlay from motion
The video docs show image-to-video through frame_images with a first_frame, so a product still becomes a short moving clip from the same picture. Pick durations from the model's supported_durations, since they differ by model. Generate one clip per angle: front, side, in hand. That also meets the multiple-angles expectation on the same page.
Whether a short clip loops cleanly depends on how you place it. This post does not claim a model produces seamless loops, so check the join in the preview before you go live.
Check a layout against the 50% rule
Before a stream, list each on-screen element with its type and the share of the frame it covers, then total the stills. The function below does that. Overlapping elements would need a different calculation, so this assumes a layout without overlaps.
def still_share(layout):
total = sum(share for kind, share in layout if kind == "still")
return total
layout = [
("video", 0.55), # presenter camera
("still", 0.30), # price card
("clip", 0.15), # product motion clip
]
share = still_share(layout)
print(f"stills cover {share:.0%}:", "ok" if share <= 0.5 else "over the 50% cap")Planning the angles
The same page asks for the product from multiple angles, so a shot list of three or four clips per product is a sensible unit: a front view, a side or detail view, and the product in use. Keep each clip on its own generation job so a bad one can be replaced without redoing the rest, and record the job ids against the stream plan. Generating before the stream is the safer route, because a LIVE has no time for a retry.
A practical set of assets
Keep the price card as a small still, which stays under the cap, and fill the larger area with generated clips of the real product. Generate from your actual product photos so the clips show the product you sell. The dynamic content post shows how to sample the first frames of a clip.
Sources
Related posts
More in Use cases
- TikTok Shop main image: no digital renderings, and the real-model shot
TikTok Shop's listing guidance excludes digital renderings from the main image and asks for a real human model image. What that means for AI product shots.
- Trick-or-treat hours video for your street, Saturday Oct 31
A 10-second street announcement for Halloween 2026: one 9:16 poster, one 6-second Wan 3.0 backdrop and a Timeline render. About $0.91 on Sume, from docs.
- Try-on with front and back garment photos: three references
When a garment's back has a print the front lacks, send shopper, front and back photos as three references. Prompt wording, a Python call and what to check.
- Tumbler or water bottle photos with AI: composite the logo in code
Brand logos on bottles drift in generated images. Make the scene with Sume's image API, then paste your transparent logo with Pillow so the mark is exact.
Written by Sume