Stella
Video editing

Create an SRT subtitle file from a video

A timed .srt file for a 73-second interview, separate from the video and checked.

Download StellaFree with the plan you already have.

A real run on 13 November 2026, on a Linux computer, using Claude Code with Opus, on a Claude plan. The waiting is sped up; it took 4 min 12 s in real time.

The short answer.

Speech recognition can write the subtitles for you. On Windows, the free Subtitle Edit has Video, Audio to text (Whisper), then saves an .srt. Whisper-based apps do the same on a Mac. Caption sites and CapCut work too, but upload the video or burn the words in. Stella (stella.sh) made an .srt for an interview with a free Whisper model running on the computer, left the video untouched, and flagged two words to check.

Stella does this on Windows, Mac and Linux.

What happened.

The .srt in about 4 minutes 10 seconds, including downloading the speech models. Checked afterwards against an answer key written before the run: a valid SRT with 19 numbered subtitles, 1.2% of words wrong (20% allowed), all 14 spoken lines shown while they're said, nothing on screen over 7 seconds, lines of 47 characters or fewer, named like the video, and the video untouched.

The interview was written for this run and voiced by two synthetic voices; the people and podcast are fictitious. The work is sped up 17×.

How Stella does it.

  1. Finds the interview video in Videos and pulls its sound out for transcription
  2. Runs the free faster-whisper speech recognition on the computer, with the large Whisper model and word-level timings, so nothing is sent to a transcription service
  3. Tries smaller models on the guest's name, which is hard to hear, and settles on its best guess
  4. Builds subtitles of at most two short lines each, timed to the words, saves Interview with Dr Okonjo.srt next to the video, then re-listens to each subtitle's stretch of audio to check it
  5. Leaves the video unchanged and names the two spots to check before uploading: the guest's first name and one phrase

The prompt.

Make an SRT subtitle file for the interview video in my Videos folder. Don't put the subtitles on the video. I just need the separate file to upload with it.

Paste it into Stella and press Enter.

This is the exact prompt from the recording. If your files are somewhere else, say where.

Do it yourself.

  1. Windows: Subtitle Edit

    Install the free Subtitle Edit and open the video. Choose Video, Audio to text (Whisper), pick a model (larger is more accurate but slower) and let it download. When it's done, check the text, then File, Save as, SubRip (.srt).

  2. Mac: a Whisper app

    macOS has no built-in subtitle maker. Free and paid apps built on OpenAI's open-source Whisper run on the Mac and export SRT; whisper.cpp does it from the Terminal with -osrt.

  3. Read it through

    Automatic transcripts get names, places and jargon wrong. Play the video with the subtitles on and fix anything odd; Subtitle Edit shows the waveform so you can nudge timings too.

  4. Keep lines short

    Aim for two lines of about 42 characters at most, on screen for 1 to 7 seconds each. Long blocks of text are hard to read and some platforms reject them.

  5. Name it like the video

    Save it as the video's name with .srt (Interview.mp4 and Interview.srt). Players like VLC load it automatically, and upload pages for YouTube or Vimeo take it as a separate caption file.

Where it stops.

  • Names and rare words are best guesses. Stella flagged the guest's first name and one phrase to check; read the file before uploading.
  • Speaker labels aren't added. Ask if you want them.
  • The speech models are big: about 5 GB downloaded, since Stella tried three sizes on the hard-to-hear name, and transcription on a processor is slow for long videos. The sound stayed on the computer; the transcript text went to the AI model you use in Stella to build and check the subtitles.
  • Recorded on Linux. On a Mac or Windows PC Stella would set up the same free speech tool, which wasn't recorded.

Questions.

How do I make an SRT file for free?

Use Subtitle Edit with Whisper on Windows, a Whisper app on a Mac, or type one by hand in a text editor for a short clip. Stella made one with a free local model.

Will it put the subtitles on my video?

No, not unless you ask. The .srt is a separate text file and the video stayed byte-for-byte the same. To burn captions into the picture, see burn subtitles into a video.

How accurate is it?

On this clear interview, 1.2% of words were wrong. Expect more errors with background noise, crosstalk or strong accents.

My site needs .vtt, not .srt. What now?

See convert SRT to VTT. For text you can read instead of captions, see transcribe voice memos to text.

Is Stella free for this?

Yes. The recording was on Stella's free plan, so the speech recognition ran on the computer with a free model and nothing was paid; it's just slower than a hosted service.

Can I control what Stella changes on my computer?

Yes. Stella works on Windows, Mac and Linux with your own files and shows each step in the chat, so you can watch it and stop it at any point. Tell it what it may not do, as the prompts on these pages do (“don't delete anything without asking”). In the Downloads recording it deleted nothing until it was told to go ahead. It also refuses commands that would wipe a whole disk, your home folder or system folders.

Checked against these sources on 13 November 2026.

Stella