How to auto-generate captions in Premiere Pro

Adobe Premiere (formerly Premiere Pro) can generate captions from your sequence's audio: transcribe the dialogue, correct the transcript, then choose Create captions from transcript. I start with the native caption track for SRT or burn-in because the same checked words can serve either delivery. Animated word styling is a separate finishing choice.

Finish the picture edit first, or get it close enough that whole phrases will stay in place. Caption blocks belong to that cut. Correct the words before creating those blocks so you don't have to repair the same name throughout a finished caption track.

Give transcription the dialogue you mean to use

Open the project and make the sequence you are delivering active. In a simple interview, the stack might read, top to bottom: C1 captions, V1 interview picture; A1 dialogue, A2 music. C1 is the result you'll create, and A1 is only an example: choose the track that carries your dialogue.

Speech to Text comes with your plan: Adobe's Generative Extend FAQ lists it among the assistive features included with a Creative Cloud membership, next to Enhance Speech and Caption Translation. English recognition ships with the app, and Adobe keeps the current set of recognition languages on its Languages supported by Speech-to-Text page.

If the sequence already has a suitable transcript, listen through it and continue with the corrections below. For a sequence that needs one, Adobe's Speech to Text sequence workflow uses these controls:

1. Choose Window > Text. In the Transcript tab, open the three-dots menu and select Generate static transcript.
2. Set Language to the language being spoken. For anything other than English, select the Download icon beside that language in the menu; the icon disappears once the pack is installed. This choice tells recognition what it's hearing; it isn't the language you want a translation to produce.
3. Under Audio analysis, select the intended audio track, or use audio clips tagged as Dialogue. With speech on A1 and music on A2, choose A1 rather than assuming the music track contains the conversation.
4. Choose whether Speaker labeling should separate the speakers. If you want the entire sequence, leave Transcribe In point to Out point only off; use it when you deliberately want just that marked range.
5. Select Transcribe. When the text appears, compare its opening words with the audio from the active sequence.

A transcript that stops partway through the sequence calls for a range check before another transcription. Wrong or absent dialogue calls for an audio-source check. Look at those two selections before treating the result as a recognition failure.

Fix the sentence before dividing it into captions

Listen against the transcript, paying particular attention to names, numbers, short negatives and punctuation. A plausible sentence can still reverse what the speaker said. The W3C's caption guidance treats automatic captions as a starting point that needs checking against the audio.

For a misspelled name, type it into the transcript's Search field and inspect the highlighted instances. Select Replace, enter the correction in Replace with, then use Replace for the selected instance. Use Replace all only after checking that every match needs the same change. These are the transcript search and replacement controls; they also let you replace a phrase with its corrected punctuation.

Read a complete thought, then listen to it. Captioning also needs meaningful non-speech sounds and speaker identification where the viewer needs them to follow the scene. Don't count on speech recognition to supply every cue.

Turn the corrected transcript into a caption track

1. Choose Window > Workspaces > Captions and Graphics.
2. In the Text panel, select Captions > Create captions from transcript.
3. For an ordinary subtitle delivery, start with Subtitle default and check the caption format. A broadcaster's specified caption standard takes priority over that starting choice.
4. Set the character length, caption duration and gaps for the pace of the speech. Choose one line or two lines. These controls divide the text into readable blocks; a short line that disappears before it can be read needs more attention than a particular character count.
5. Select Create captions. The captions appear on a separate timeline track.

Adobe's caption creation controls provide those segmentation choices. Read each block as a phrase rather than cutting a sentence into equal visual lengths. Keep related words together when the speech gives you room.

Check the words at the playhead

Move the playhead a little before a caption begins and play through its end. Does the text arrive with the phrase, remain readable and leave before a different speaker takes over? Adjust the caption segment's start and end handles against the speech and waveform, as Adobe describes for timing subtitles on the caption track.

Now play across the picture cut. A caption may correctly continue across it if the same sentence continues; a new shot alone is not a reason to cut the words. Listen for the phrase you are captioning and watch for a block that hangs over unrelated dialogue.

If picture timing changes later, check the affected caption blocks and the phrases after the edit point again. Adobe recommends captioning when the cut is finished or nearly finished; its caption workflow notes also warn that deleted caption segments need manual recreation if restored footage needs them. Restoring the picture won't bring those words back on its own.

Translate the checked captions into another track

Caption translation uses a cloud service with third-party translation models, including Google Translate and Microsoft Translator. Confirm that cloud processing suits the material before using it. Translation changes the caption language; it does not replace the spoken audio.

In the Text panel's Captions tab, select Translate captions, choose a target offered in the language menu and select Translate. Premiere creates a new caption track. Adobe's caption translation workflow keeps the language versions on separate tracks, with eye icons to show or hide them.

Have the translated wording checked for meaning, names and readable phrase breaks before delivery. To isolate a language track, Option-click its eye icon on macOS or Alt-click it on Windows; clicking it again with the same modifier mutes that track. Then enable the language you intend to export. Seeing two languages in the Program Monitor is a reason to check both track visibility and export selection.

Choose SRT or words fixed into the picture

Select Export at the top of Premiere and open Captions. For a separate subtitle file, choose Create Sidecar File and SubRip Subtitle Format (.srt). For words that must remain visible in the video, choose Burn Captions Into Video. Those caption export options serve different deliveries: burned-in captions cannot be switched off during playback.

Inspect Captions > Caption Tracks before exporting. SRT carries one caption stream per file. If two or more caption tracks are enabled, Premiere selects the lowest-numbered track by default. Check the selected language track explicitly and export a separate SRT for each language required. With burn-in, multiple active caption tracks can put multiple languages into the picture.

If the Captions section is disabled, check that the sequence contains a caption track. A transcript alone is not that track. Embedded broadcast captions, manual caption entry and imported subtitle files belong to the broader captions and subtitles workflow; the captions hub keeps those finishing tasks together.

Open the exported video in its destination player and load the matching SRT if that is the delivery. Check the opening phrase, a phrase after an edit and the final line, then watch the full piece for wording and timing. For burn-in, inspect the words at the intended viewing size and adjust the caption font or size if they are hard to read.