TikTok auto captions or burned-in captions on an AI video?
Use both when it matters: TikTok's auto captions are a viewer toggle that needs speech; burned-in captions from Sume ($0.20 per clip up to 60 s) always show.

TikTok's auto captions are a feature viewers switch on; they come from speech, so they do nothing for a silent or music-only AI clip. Burned-in captions are pixels in the MP4, so they show for everyone and carry your wording and style. For a finished AI video, burn captions with Sume ($0.20 per clip up to 60 seconds) and let TikTok's own captions cover the rest.
TikTok's 2022 newsroom post describes the two features: auto-generated closed captions that creators and viewers can use, and translations of captions, descriptions, and text stickers. It says they were in early rollout with an initial set of nine languages: English, Portuguese, German, Indonesian, Italian, Korean, Mandarin, Spanish, and Turkish. That list is four years old, so check what the app offers now.
Where each one wins
| Question | TikTok auto captions | Burned-in (Sume) |
|---|---|---|
| Who controls it | The viewer turns it on | You, in the file |
| Needs speech | Yes | Not with cues |
| Style | TikTok's | slam, punch, tiktok-green, and Hangul styles |
| Wording | Machine transcript | script_text aligns your exact script |
| Can be switched off | Viewer's choice | No |
| Cost | Not stated on the cited page | $0.20 per standalone job, clips up to 60 s |
Burned-in captions on Sume
Video captions takes a public HTTPS video_url and burns captions in. If you omit style, Latin text defaults to slam and Korean to black-outline. tiktok-green is a named style. language is only a speech-to-text hint and never selects the style or font.
If you give script_text, Sume keeps the STT word timings and aligns your text to them, and the job can fail with script_alignment_mismatch if the script and speech disagree. Send source_caption_id instead of video_url to restyle without transcribing twice.
A practical split
Burn captions when the clip depends on the words: dialogue, a price, a disclosure line. Rely on TikTok's captions for accessibility and translation, since viewers can read them in their own language. Make sure the burned text sits clear of the app's buttons; the anchor-ratio post explains how to place lines.
A burned-in AI disclosure also helps on the label side: TikTok's 2023 label post describes the toggle, and the Shop pages accept an on-screen note as an alternative way to disclose.
Cost and effort
Burning captions on every clip adds $0.20 per job for clips up to 60 seconds, so a month of 30 posts adds $6. That is small next to the render cost, but it is not free, and you can skip it for clips where TikTok's captions will do. If you restyle a clip, use source_caption_id so you do not pay to transcribe it again.
Check one more thing before you burn: if the language is Korean, use a Hangul style, since Korean on a Latin style returns a 400 error.
Check the result on a phone, not only on a desktop preview. Captions that fit a 16:9 window can sit under the TikTok caption, username, or buttons on a 9:16 screen. Keep the burned text in the middle band of the frame and re-render if a line is covered.
Sources
Related posts
More in Media tools
- TikTok horizontal ad minimum is 960x540: which 16:9 sizes clear it
TikTok's in-feed page lists 960 x 540 as the minimum for 16:9. Sume Timeline accepts 960x540, 1280x720 and 1920x1080, all at $0.10 per output minute.
- TikTok non-Spark ad captions: no links, @ or hashtags, lint first
TikTok's auction page says non-Spark captions allow no links, @ symbols, or hashtags; Spark Ads take 4 caption lines. Lint the text in Python before upload.
- TikTok 'Reference to Video' vs Sume frame_images for product moments
TikTok's Symphony Creative Studio lets advertisers prompt exact images and products at set moments. How the job maps to Sume frame_images and references.
- White backgrounds hide TikTok's ad UI: dim the clip, check for free
TikTok's reservation page warns that white or transparent backgrounds can hide the nickname and captions. Dim with video-filter ops, check first, then re-probe.
Written by Sume