How to Add Captions in Descript

Most “add captions in Descript” posts skip the split that actually decides the export. Captions that stay on the picture and subtitles you upload beside the file are two products. Official help says so. One is a layer you style. The other is an .srt or .vtt from Export → Subtitles. This guide is the working path for both — composition, transcript, names, timing, style, export — using the one editor we send for this brief: Descript.

When Descript captions are the job — and when you should package instead

Use Descript when the words already live in a composition. A video podcast, a talking-head YouTube cut, an interview you radio-edited: the script is the caption source. Apply a style, correct the guest’s name, drag a late word on the Wordbar, then decide burned-in or sidecar. That is the caption pass.

Skip this pass if the remaining job is the radio edit. Deleting the tangent at 18:40 is a transcript job, not a font job. That working path is How to edit a podcast in Descript — an ordinary guide link, not a second money hop. Skip both if the file is already a 9:16 keeper and you need a packager’s word-by-word house style. That path is How to add AI captions to YouTube Shorts. Finding 5–15 Instagram clips inside a finished hour is How to make Instagram clips with Klap. Klap and the Shorts packagers stay foils on this page, not Try buttons.

The seven-step Descript caption pass

1. 01

   Confirm the file is yours, then open or import a composition — not a feed enclosure

   This pass is for a podcast or video you hosted, filmed, or have written permission to caption. Guest likeness and a licensed bed are not automatically yours to republish with type on the picture. Then get a real file: a mixed MP4, a talking-head take, or an episode you already radio-edited. Open the composition that holds that cut, or start a project and import from your computer, from Zoom if that is the source, or from a phone upload. You can also record straight into a composition. An RSS XML feed is not a caption input.

2. 02

   Wait for the transcript — captions are generated from the script, not a separate timer

   Transcription starts after the file lands. Official caption help is explicit: Descript generates captions from your script and syncs them to the timeline. You do not draw boxes on a waveform to invent a track. Wait until the script is there. Every uploaded or recorded file counts against media minutes. If the words are still a ramble you have not cut, stop and do the radio edit first — captions will faithfully print the tangent.

3. 03

   Add a captions layer from the Captions panel — pick All scenes or This scene first

   Open the right-hand sidebar and click the Captions panel. Official help: use the toggle at the top to apply captions to just the current scene or all scenes, and do that before you apply a style. Then select a captions style. That adds a new captions layer. Descript builds the on-screen words from the script. A text layer is a different object — it is not connected to the transcript unless you convert it. If nothing appears, you have not applied a style yet.

4. 04

   Fix names and speaker labels in the script — the captions follow those words

   Proper nouns, product names, and guest spellings fail first. Official Correct your transcript help: double-click the word, click Correct (or press C), type the right spelling, then Correct this instance or Correct All. For speakers, type @ in the script, or select text and Change speaker, then create or rename the label. Automatic speaker detection can run on import; identify each voice once. Captions update with the script. Do not restyle a track that still says the guest’s name wrong.

5. 05

   Fix timing on the Wordbar when a card sits late or early

   If a word is on screen before it is spoken, or hangs after the mouth has moved on, the alignment is wrong — not the font. Official Wordbar help: zoom the Timeline until the word is ungrouped, hold Command (macOS) or Control (Windows), and drag the word’s start or end boundary. That same alignment drives captions and exported subtitles. Dragging a word left shortens a gap; dragging right adds a gap clip or a freeze frame. We describe that action. We are not inventing a “nudge captions 200ms” knob that public docs do not name.

6. 06

   Style for YouTube or for social — same layer, different job

   Select the captions layer. Open the Properties panel. Official Style controls: font, size, and alignment; text and background colors; borders; styles for active words and future words. For a YouTube episode people will watch with sound on, keep type readable and out of faces — a clean card, not a shouty highlight reel. For Shorts, Reels, or TikTok, use the active-word / future-word styles so muted viewers can follow one word at a time, then move the layer out of the platform chrome. For two hosts, official help: a separate captions layer per speaker, then pick the speaker from the layer options dropdown. Each layer can be styled and positioned on its own.

7. 07

   Export burned-in video, or a sidecar SRT / VTT — then upload the matching pair

   Official help treats these as two products. Captions stay on the picture when you share or upload the video. Subtitles are a separate file. For burned-in type, leave the captions layer on and Export a local video (MP4). For toggleable YouTube captions, Export → Subtitles tab and pick .srt or .vtt. Published subtitle settings: Show speakers, Word level timestamps on .vtt (karaoke-style), Max characters per line, and Max lines per card. Upload the video without burned-in type if you only want the sidecar, then add the file in YouTube Studio. Do not burn the track in and also upload the same words unless you meant a double stack.

Burned-in vs sidecar vs handing the clip to a packager

“Add captions in Descript” is a search. The decision is whether the type has to stay on the picture. The Descript columns match the published Captions panel, Properties / Style controls, and Export → Subtitles page as of August 2026. Packagers stay qualitative here — we already wrote that desk, and this page does not hop there.

Open the composition — then wait for the script

Official getting-started copy gives you two starts: record in Descript, or import an existing file. Import paths they publish include upload from your computer, import from Zoom, and upload from a phone. If the episode already lives in a project, open that composition. Do not start a second binder “just for captions” and re-import the same hour — media minutes count every file.

Bring the cut you will actually publish. A radio-edited video podcast is the right input. A talking-head YouTube file is the right input. A raw two-hour ramble will caption the ramble. If you still need the radio edit, do that first on How to edit a podcast in Descript. If the take still sounds like a kitchen, clean it first on How to clean up podcast audio in Descript. Those are ordinary site paths. This page already hops to Descript.

Transcription is the next automatic step. Official caption help does not ask you to type a track from scratch. It generates captions from the script. Wait. Then read the first minute. If the script is empty, the Captions panel has nothing to style.

Generate the captions layer — All scenes or This scene

Open the right-hand sidebar. Click Captions. Official Add and style captions help: use the toggle at the top to apply captions to the current scene or all scenes, and do that before you apply a style. Stock-media help uses the same split — All scenes for the whole video, This scene for the active one. Then select a style. That creates the captions layer and syncs it to the timeline.

Captions follow the script. Delete a sentence in the document and the on-screen words go with it. Official help is also explicit about the other object: a text layer is not connected to the transcript. You can convert a text layer to captions if that is the surface you started on. We describe that action. We are not inventing a third “auto-caption wizard” beyond the Captions panel.

Two hosts who should not share one card: official help adds a captions layer per speaker. Select the layer, open Properties, use the options control next to the layer type, and choose the speaker. Repeat. Each layer can be styled and positioned on its own. That is the published multi-speaker caption path — not a color we guessed.

Fix names, then timing — script first, Wordbar second

The captions will print whatever the script says. Official Correct your transcript help: double-click the word (or the adjacent word), click Correct in the hover toolbar or press C, type the right word, then Correct this instance or Correct All. Correct mode — Option+C on Mac, Alt+C on Windows — lets you click and type without opening the toolbar each time. After a correction, Descript re-checks alignment; a dotted underline is the published “still aligning” state. Hold Z and click to cycle capitalization. Hold X and click to cycle punctuation.

Speakers are labels, not decorations. Official Speakers help: type @, or select text and Change speaker, then create a name or pick an existing one. Automatic detection can run on import; Identify speakers is the published follow-up. Rename a label from any instance. If the guest is attached to the wrong paragraph, reposition the speaker label in the script. Show speakers on a sidecar export is a later checkbox — it only helps if the labels are right.

Timing is a Wordbar job when a word is misaligned. Official copy: the Wordbar is how you adjust a word’s start and end so the transcript, captions, and exported subtitles line up. Zoom the Timeline until the word is ungrouped. Hold Command (macOS) or Control (Windows) until the trim handle appears, then drag that boundary. Word-level only. If the card is late because the script still includes a pause you meant to cut, that is a radio-edit, not a caption nudge.

Style for YouTube or social — then pick the export

Select the captions layer. Open the Properties panel. Official Style list: font, size, and alignment; text and background colors; borders; styles for active words and future words. That last pair is how a word-by-word highlight happens. We are not going to invent a toolbar button named Karaoke. If a third-party write-up uses that word, the published controls are still active / future words.

For a YouTube episode people watch with the sound on, keep the card readable and out of faces. A dark background behind light type is the usual accessibility move; confirm it on your picture. Leave active-word fireworks for the vertical. For Shorts, Reels, or TikTok, use the active-word style so a muted phone can follow, then drag the layer off the username bar and the engagement stack. Official help does not publish a “safe zone” preset. We describe the action: move the layer until a phone preview does not cover the words.

Layouts can reuse a caption look across scenes and compositions — official caption help points at custom layouts for that. Save the look you would actually ship, not the first style you clicked. Then decide the file that leaves the app.

Burned-in: leave the captions layer on, Export, local video, MP4. Official help: captions that always stay on the video, even when shared or uploaded. That is the social file. Sidecar: Export → Subtitles tab. Published formats are .srt and .vtt. Published settings are Show speakers, Word level timestamps on .vtt, Max characters per line, and Max lines per card. YouTube Studio will take the sidecar on a long upload. Official YouTube publish from Descript’s web app still does not list Shorts as a destination — if the cut is a Short, download the vertical and upload it in Studio.

Do not burn the track in and upload the same words as a sidecar unless you want viewers to read the sentence twice. Do not publish a watermarked Free-plan 720p export to a feed you already monetize. Sit on a paid plan before the captioned file is the one subscribers get. Confirm checkout on descript.com/pricing. We do not print a commission rate.

When a packager or a clipper is the honest next hop

Stay in Descript when the composition is the show and you need accurate type — burned-in for a muted vertical, or a sidecar for YouTube. Hand the file off when the remaining job is a packager’s house style on a clip you already chose. That pick lives on How to add AI captions to YouTube Shorts and Best AI caption tools. Neither is a hop from this page.

Finding 5–15 Instagram clips inside a finished episode is a clipper job: How to make Instagram clips with Klap. A 2–5 minute narration you still have to generate is How to make a YouTube voiceover with ElevenLabs. The radio edit and the cleanup remain the two Descript pages above. None of those names are Try buttons here. The only hop on this page is Descript.

We do not print an affiliate commission rate, and we do not invent a “this will earn you $X” line. The product question is whether the on-screen words match the script you already trust — and whether you exported the file the destination can actually use.