Square TikTok ad from an avatar video: 1:1 at 720p vs 640 px minimum
TikTok lists 640 by 640 as the minimum for square non-Spark video. Sume avatar video supports 1:1 at 720p, which is above that floor.
TikTok's In-Feed page lists 640 by 640 as the minimum for square non-Spark video. Sume's avatar video supports aspect_ratio of 1:1 and renders at 720p, so a square take is above that minimum. TikTok recommends 9:16, so test the square against a vertical version before you commit budget.
What TikTok's page says
On "TikTok Auction In-Feed Ads" (read 2026-10-05), non-Spark sizes are vertical 9:16 at 540 by 960 or more (recommended), horizontal 16:9 at 960 by 540 or more, and square 1:1 at 640 by 640 or more.
What avatar video offers
Per the avatar video docs, aspect_ratio accepts 1:1, 3:4, 9:16 (the default), 4:3 and 16:9. The output is 720p only, and the duration is 4 to 60 seconds. A 1:1 frame at 720p is 720 by 720, assuming the output matches the stated 720p on its short side; confirm with video inspect.
curl -s https://api.sume.com/v1/catalog \
-H "Authorization: Bearer $SUME_API_KEY" | head -c 600Square or vertical
TikTok marks vertical as the recommended shape on that page. A square gives less height to work with, so a caption line and a face compete for space. If you run both, render both and keep the winner.
- Ask for
1:1for the square test and9:16for the control. - Keep the script the same for a fair comparison.
- Check size, length and bitrate against the page before upload.
Captions
Burn captions after the avatar render with video captions. The docs say inline captions on avatar video are rejected above 60 seconds of estimated length, so keep the take short or caption in a second pass.
Sources
Related posts
More in Sume Avatar 1.0
- Griffin-Lite 26 of 54 Turing result: what the sample size says
Tavus reports 26 of 54 callers fooled by Griffin-Lite after a one-minute call. A 95% interval is about 35% to 61%. Python computes it, and says what to claim.
- Tavus Video to Face replica vs a Sume avatar from a photo or prompt
Tavus builds a replica from video or a photo; Sume builds an avatar from a prompt, props or a public photo URL. What each input gives you, and what it costs.
- What an AI avatar may not claim on TikTok Shop
TikTok Shop prohibits AI that impersonates real people or invents doctors and experts to endorse products. An avatar can present, but it can't be a fake expert.
- UGC hook, demo, CTA: one avatar video with a silence beat
Build a 9:16 UGC-style ad as three scenes in one avatar talking-video request: spoken hook, silent demo beat, spoken CTA, inside the 4-60 second window.
Written by Sume