How to transcribe a video automatically in Premiere Pro and get word-by-word captions

Step by step

1. Open the transcription — In the Text step of the panel, click + Add text: the window opens on the ✨ AI transcription tab.

2. Set the language — In Spoken language, keep Auto-detect or pick the language spoken in the video.

3. Pick the voice track — Under Audio tracks, tick the track or tracks that hold the voice, or tick All non-muted tracks.

4. Choose what to transcribe — In What to transcribe, keep Between in / out points (set them with I and O in the Premiere Pro timeline) or choose Whole sequence.

5. Check the cost and start — Click ✨ Start transcription, check the Duration to transcribe and the Credit cost, then click 🚀 Start.

6. Check your lines — The window closes on its own and your lines appear in the Text step, split into blocks according to your preset.

What does AI transcription do?

AI transcription listens to the audio of your active sequence in Premiere Pro and writes the captions for you, with a timecode for every word. Each word appears on screen exactly when it is spoken, which is what word-by-word animated captions, karaoke-style active words and synced sounds rely on.

It all happens inside the panel: no audio file to export, no text to copy back. When it’s done, your lines are in the Text step, already split into blocks and laid out by your preset.

Why it matters: typing and timing captions by hand takes far longer than the video itself. Here you check and fix a ready-made text instead of building it word by word — and every word is already in the right place on the timeline.

How to start a transcription

1. Open your editing sequence in Premiere Pro and click in its timeline so it is the active one.
2. In the panel, go to the 2 Text step and click + Add text. The window opens on the ✨ AI transcription tab (the other tabs are 📥 Import and ✍ Write).
3. Set Spoken language, Audio tracks and What to transcribe (details below).
4. Click ✨ Start transcription. A summary window, Start transcription?, shows the scope, the tracks, the language, the Duration to transcribe, the Credit cost and your Credits remaining.
5. Click 🚀 Start. The panel shows Extracting audio…, then AI transcription… (about 1 min per 10 min of audio) with a running timer and a Cancel button.
6. When it’s finished, the window closes on its own and Transcription done: N subtitles. appears.

No credits, or no voice at all? The bottom of the tab offers Import a .srt / from the project and Type by hand: see Import an SRT file or your Premiere captions and Write your own text.

Which settings should I choose?

Option	Where	What it does	When to use it
Spoken language	✨ AI transcription tab	Auto-detect, or one of 26 languages. Your choice is remembered for the next time and shown in the summary window.	Pick the language when part of the text comes out in the wrong language (a strong accent, for example).
Audio tracks	Same tab	One box per audio track of the sequence; empty or muted tracks are flagged (empty) or (muted). All non-muted tracks sends every track that isn’t muted.	Tick only the voice track when music or effects sit on other tracks.
What to transcribe	Same tab	Between in / out points (default) or Whole sequence.	In/Out to transcribe one part, or to save credits.
Block splitting note	Same tab	Block splitting: per your preset — click here to adjust it (Style › 🔧 Layout).	Click it to change how the lines are cut before you transcribe.

Which languages are in the list?

Auto-detect recognizes more than 100 languages. You can also force one of these 26, each written in its own language in the menu: French, English, Spanish, Italian, Hindi, हिन्दी + English (Hindi and English mixed in the same sentences), Portuguese, Indonesian, Turkish, Vietnamese, Russian, Japanese, Thai, German, Dutch, Polish, Ukrainian, Arabic, Chinese, Korean, Tagalog, Tamil, Telugu, Bengali, Marathi and Urdu.

How do I set In and Out points in Premiere Pro?

In the Premiere Pro timeline, move the playhead to where the part starts and press I, then to where it ends and press O. With Between in / out points, only that range is transcribed — one point is enough: with only an In point, the panel transcribes from there to the end; with no point at all, it takes the whole sequence.

Tip: the voice is often on one track only. Ticking just that track gives a cleaner text than sending the music along with it — and if you pick a track that is silent in the range, the panel tells you before spending anything: None of the selected tracks has audio in the range to transcribe.

How are the lines split into blocks?

The transcription is cut into caption blocks with the settings of your preset, in Style › 🔧 Layout › ✂ Subtitle splitting:

• Characters per block (2 to 200): the maximum length of a block.
• Split if silence > (s): a pause in the voice longer than this starts a new block (transcription only).
• Cut at punctuation (. , ! ? …): blocks end at the right place in the sentence.
• Attach isolated punctuation to the word: a lone “!” is glued to the word before it.

A block never runs across a cut in your edit: the clip boundaries on the transcribed tracks are hard breaks. Want a different split afterwards? ✂ Re-split in the Text step recuts everything without touching your preset — see Edit lines, blocks and timecodes.

How much does a transcription cost?

AI transcription uses credits, counted in seconds of audio: 1 credit = 1 second. The window shows them as a duration in Credits remaining (for example 2 h 41 / 3 h 00); on a subscription, its renewal date is shown underneath.

Plan	AI transcription included
Free	10 minutes to try it
Pro	3 hours per month
Max	15 hours per month

• The Credit cost in the summary is about the length of the range you send.
• If you have offered credits (shown as Offered credits (no expiry)), they are used first.
• If your credits don’t cover the duration, ⚠ Not enough credits for this duration. appears and 💳 Change plan replaces 🚀 Start. Shorten the range with In/Out points, or see Free, Pro and Max.
• A transcription that finds no speech is still charged: check your track before starting.
• Generation itself never uses credits.

Note: your audio goes to our transcription service only to be turned into text; once the text is delivered, we ask the service to delete it. Details in our privacy policy.

Long videos, waiting time and interruptions

• Length: one transcription can send about 1 h 20 of audio. Beyond that, Audio too large to upload appears with the maximum duration. For a longer video, split the edit into several sequences of less than 1 h 20 each and caption them one by one: the panel keeps a separate text for each sequence. Note that a new transcription replaces the text of the current sequence in the panel (Ctrl+Z / Cmd+Z brings the previous one back).
• Time: count about 1 minute of processing for 10 minutes of audio. If it takes longer, the panel says The AI is taking longer than usual — we keep waiting, nothing is lost.
• Timeout: if the service still doesn’t answer, run the same transcription again without changing anything (same In/Out points, same tracks, same language): the job already started is resumed, at no extra credits.
• Changing sequence: if you switch sequences while it runs, the result is not applied, so nothing is overwritten. Go back to the original sequence and run it again.

More error messages and their fixes are in Troubleshooting.

What happens after the transcription?

1. Your lines appear in the Text step, block by block, with their timecodes.
2. The Compo Auto of your preset is applied right away: important words get their style and their place, as set in your preset.
3. Each word is anchored to your clips: if you cut or move clips afterwards, the captions follow — see Captions that follow your edits.
4. Check the text, fix a word with a click, then move on to layout and generation.

Tip: cutting your silences first with Autocut means less audio to send — and fewer credits used.

Frequently asked questions

Can Premiere Pro make word-by-word animated captions automatically?

Yes, with Subkick Caption. Its AI transcription writes your captions from the audio of your sequence, with a timecode for every word, then the plugin animates them word by word in your timeline. In the panel, open the Text step and click + Add text › ✨ AI transcription.

Which languages can Subkick Caption transcribe?

Auto-detect recognizes more than 100 languages, and 26 can be picked directly in the Spoken language list, including English, Spanish, French, Hindi, Arabic, Japanese and Tagalog. A Hindi + English entry handles speech that mixes both languages.

How much AI transcription is included in Subkick Caption?

Subkick Caption includes 10 minutes of AI transcription on the Free plan, 3 hours per month on Pro and 15 hours per month on Max. Credits are counted in seconds of audio, and the confirmation window shows the cost before anything starts.

How long can a video be for one transcription?

One Subkick Caption transcription can send about 1 h 20 of audio. For a longer video, split the edit into several sequences and caption them one by one: the panel keeps a separate text for each sequence. Count about 1 minute of processing for 10 minutes of audio.

Do I pay twice if the transcription takes too long?

No. If the transcription service doesn't answer in time, run the same transcription again without changing anything (same In/Out points, same tracks, same language): Subkick Caption resumes the job already started, with no extra credits.

Is my audio kept after the transcription?

Subkick Caption sends the audio of the chosen range to our transcription service only to turn it into text. Once the text is delivered, we ask the service to delete the audio. Details are in our privacy policy.