Captions, follow the speaker and auto-zoom

Davnai can write the words on the video, timed to your voice, from the recording's transcript. They are part of the picture ("burned in"), so they show on every platform, even with the sound off.

Where: Make Shorts and Split into parts (on by default); Make the edited video on the review screen, Trim, Resize and Join (off by default; in Join, clips without a transcript simply have no captions); and Add captions (Video Studio tile, or ⋯ → Add captions on a recording), which makes a captioned copy of the whole recording. Captions need a transcript; windows without one say so. Switch Captions on or off in the window.

Styles: 24, in five groups (each tile shows a sample; NEW marks the ones added with the library):

• Clean: Clean, Subtle, Cinema, Serif, Minimal.
• Bold: Bold, Impact, Headline, Punch.
• Karaoke (the word being said lights up): Highlight, Sweep (the colour fills each word as it's said), Glow, Underline.
• Boxed: Boxed, Pill, Sticker, Marker.
• Creator: Word pop (one big word at a time), Bounce, Comic, Two-tone, Handwritten, Typewriter, Big words.

Make a style your own:

• Font: 20 fonts (Montserrat, Anton, Bebas Neue, Bangers, Luckiest Guy, Permanent Marker, Caveat, Playfair Display and more), or keep the style's own. The finished video uses exactly the same font as the preview.
• Word said: the colour of the word being spoken. Your Brand Kit colours show first.
• Words: Auto, or 1, 2, 3 or 4-6 words per line.
• In: how each line comes in: the style's own, Pop, Bounce, Fade, Type or None.
• Bigger key words: numbers, money, names and strong words ("never", "free", "secret") are bigger and in the word colour.

Place: Bottom (above the platform buttons on vertical videos), Middle, Top, or Follows your voice: captions jump to the top and grow when you raise your voice, sit in the middle for a question, and stay low when you're calm. They move at most every few seconds so they don't jump around. Follows your voice works from the recording's sound; for a part with no sound wave yet, captions stay at the bottom. A few words show at a time (by default up to 3 on vertical videos, up to 7 on landscape), split at pauses and at the end of sentences.

The editor's live preview shows the captions as they will be made.

On Finish my video, the Finish step has a Captions section, off by default. Switch it on to make the edited video with captions. For a video you already made without captions, choose Make again with Captions on.

Choose Caption language: As spoken (the transcript's own language) or one of 20 languages to translate into: English, French, Spanish, Portuguese, German, Italian, Dutch, Arabic, Hindi, Chinese (Simplified), Japanese, Korean, Russian, Turkish, Indonesian, Swahili, Yoruba, Hausa, Igbo and Nigerian Pidgin. For example, speak English and show French captions. Davnai translates the transcript phrase by phrase with AI; each translated phrase shows while the original phrase is spoken (word highlighting inside a phrase is spread evenly, so it's close rather than exact). The first time a recording is translated into a language takes a little longer (usually under a minute); after that the translation is reused for every Short, part or copy. The editor's live preview shows the translated captions too, as the made video will. Translations can be imperfect, especially for less common languages: check before publishing. Captions work for any spoken language the transcript supports, including non-Latin scripts (Arabic, Hindi, Chinese, Japanese, Korean).

Captions use the transcript's words, so fix the word once and every caption follows. In the video editor, pause the video with Captions open and click the word on the video, or double-click it in Words; type the right spelling and choose this one or all of them in the video. Brand Kit → Your words keeps names and brand words spelled your way on every new recording. If all the words are wrong, the transcript was probably made in the wrong language: open the transcript and use Wrong language? Transcribe as…. Captions shown in another language can be fixed the same way: pause with Captions open and click the word; the fix goes into the translation (it isn't translated again, so it costs nothing more). Names fixed in the spoken words carry into the translation by themselves.

Captions need a transcript with word timings. Recordings without speech (or without a transcript) have no captions; the video is still made.

When a landscape video becomes vertical or square, Follow the speaker finds the face and keeps it in frame as the person moves, like a camera operator: it holds still for small movements and pans smoothly for bigger ones. It's in Make Shorts, Split into parts (when the shape changes) and Resize. Best for one person talking. If a face is seen only now and then, the frame holds still where the face was; if no face is found at all (a screen recording, or the person is very far away), the frame stays in the middle and the finished video's card says No face found · centred. For several people talking, Blurred background keeps everyone in view. On a new shot (a cut to another angle or person) the frame moves straight to the face instead of panning across.

When pauses and mistakes are removed, the video jumps at every cut. Auto-zoom punches in slightly (about 12%) on every other part, centred on the face, so the jumps look intentional. It's in Make Shorts and Split into parts (needs "Remove pauses and mistakes") and in Make the edited video when the edit has cuts.

It never counts toward your video minutes. Following the speaker and auto-zoom add a quick pass to find faces, so the estimated time is a little longer.