Do I get subtitle files as well as burned-in captions?

Yes. Every render produces SRT and VTT files alongside the video, generated from the same word-level timing as the burned-in captions.

How accurate is the timing?

It comes from running the finished audio through speech recognition, so it reflects what was actually said rather than an estimate based on how many characters the script had.

Do captions work on vertical video?

Yes, and they adjust for it: on 9:16 the text is larger and moves to the centre of frame, so platform interface elements along the bottom edge do not cover it.

Can I caption a video I made elsewhere?

No. Captions are produced as part of generating a video with Scenes AI — there is no path for uploading finished footage to be captioned.

AI captions

AI captions in 26 animated styles

Every Scenes AI video can have word-level animated captions burned in, in any of 26 styles — and the style is chosen automatically to suit the format, so a Short gets punchy word-by-word captions while a story gets something calmer. Each render also produces SRT and VTT subtitle files you can upload separately.

Twenty-six styles, picked for you

You can choose one, but you do not have to — the default follows the kind of video you are making.

Styles include tiktok, hormozi, karaoke, cinematic, typewriter, neon, glow, fire, lightning, bounce, wave, shake, glitch, explosive, 3d and gradient.
YouTube Shorts default to a bold word-highlight style; Reels default to karaoke; stories default to a restrained cinematic style; poems fade.
Font, size, colour, highlight colour and position are all adjustable if the default is not what you want.
On vertical video the caption size increases and the position moves to centre automatically, because bottom-anchored captions get covered by platform UI.

Timing that actually matches the speech

Word-level timing comes from running the finished audio through speech recognition, not from estimating durations from character counts. That applies to narration generated by a speech engine and to dialogue the video model produced itself, so lip-synced scenes get accurate captions too. Captions are paged rather than dumped on screen — six words at a time by default.

Subtitle files, not just burned-in text

Burned-in captions are right for social. Uploaded video usually wants a sidecar file as well.

Every render produces both SRT and VTT files alongside the video.
Both use the same word-level timing as the burned-in captions, so they cannot disagree.
Useful where captions must be toggleable, or where a platform requires a subtitle file for accessibility compliance.

Captions in the language you chose

The language you pick governs everything the audience hears and reads, captions included. Where a scene contains quoted non-English dialogue, the video engine is explicitly told the language so it does not translate it back to English — which would leave the captions and the audio saying different things.

FAQ

AI captions questions

Can't find what you're looking for? Visit help and support or email hello@usescenes.com and a human will reply.

Yes. Captions are a setting like any other, and you can also change the style, font, size, colour and position rather than turning them off entirely.

Your first video starts with one sentence.

Describe what you want. Scenes AI writes the script, generates visuals, adds voiceover, and renders the final cut. You approve at every step, and nothing expensive runs until you have.

Create Your First Video

Pay as you go · No subscription · Export up to 4K