How do I get subtitles and SRT files for my videos?

This page may contain affiliate links from which Blur Studio earns commissions.

A meaningful share of video views happen with the sound off, and viewers without captions on a video that needs them tend to leave within seconds. Subtitles aren't a nice-to-have anymore for anything meant to be watched on a phone, in an office, or by someone who needs captions to follow along at all.

The free option first

If your video is already going on YouTube, its automatic captions are free and reasonably good for clear, ordinary speech in common languages. For a lot of general-purpose videos, that's genuinely enough, and there's no reason to pay for something you already have for free. Where it falls down is accents, background noise, technical vocabulary, and anything with brand names, product names or numbers, where auto-captions misfire more often than people expect. If your video is going on social platforms instead of YouTube, several video editing apps built for that workflow (CapCut and similar tools) also generate free auto-captions styled for vertical video, which is often the faster route if you're already editing in one of them rather than adding a separate transcription step. For a live webinar instead of a recording, adding live captions to webinars and calls is a different setup.

When you need something better, or your video isn't on YouTube

If your video lives on your own site, in a training platform, or anywhere else without YouTube's free captioning, you need a separate transcription step. ElevenLabs' Speech to Text (Scribe v2) transcribes the audio track and returns word-level timestamps, which is what an SRT file actually needs: not just the words, but exactly when each line should appear and disappear. Pricing through the API is $0.22 an hour of audio, and it handles files up to 3GB or 10 hours long, so a single long training video or webinar recording is no problem in one pass.

The output from a transcription API isn't automatically an .srt file; you (or whatever tool you're using) need to format the timestamped text into SRT's specific structure, which most video editors and captioning tools can do once you feed them the timestamped transcript. If you're not comfortable with that formatting step yourself, a tool that wraps the Scribe v2 API and exports SRT directly, or a captioning service, saves you that part. If you catch a wrong word or name after generating the transcript, a change ElevenLabs made on 28 September 2026 added transcript editing for both batch and real-time transcription, so you can fix it directly in the text before you format the final SRT file, rather than starting the transcription over. If you already have the script, lining up subtitles with a script skips transcription entirely.

What makes a caption file actually readable, not just accurate

Getting the words right is only half the job. A technically accurate SRT file can still be hard to follow if the lines are too long or stay on screen too briefly. The guidelines most professional captioners work to: roughly 32 to 42 characters per line, a maximum of two lines on screen at once, and a minimum display time of about one second per caption, so a viewer actually has time to read it before it disappears. A raw transcription API only gives you word-level timing, when each word was spoken, not any of these readability rules. Whatever tool turns that timed transcript into an SRT file needs to handle line-breaking and minimum display time on its own, or you'll end up with captions that are technically correct but flash past too quickly for anyone to actually read.

A worked example: a 12-minute training video

A single 12-minute training video costs about $0.044 to transcribe through Scribe v2's API, genuinely trivial even once you add the formatting step. The real cost on a video that length isn't the transcription, it's the review: checking every proper noun and number by watching with captions on takes roughly the 12 minutes it takes to watch the video once, plus a few extra minutes to fix anything wrong. For a library of 40 training videos at that length, that adds up to a meaningful chunk of a day's work, which is exactly the kind of one-time cost worth doing properly rather than rushing, since the resulting captions get reused every time someone watches.

What I'd actually tell someone making training or marketing videos

For internal training videos, where the audience is forgiving and the content changes often, free auto-captions reviewed quickly by a person are usually the right amount of effort. For anything client-facing, a product demo, a marketing video, a webinar you're reusing for months, I'd spend the extra bit of time on a proper transcription pass and a careful human check, because a wrong product name or price in a caption looks sloppy in a way viewers notice even if they can't quite say why. The cost difference between "good enough" and "actually accurate" here is small; it's mostly a matter of whether anyone reviews the output. There's also a real accessibility angle for public-facing video: captions increasingly factor into accessibility standards for organizations with any kind of compliance obligation, and "we have some captions" isn't automatically the same as meeting that bar. If your video is genuinely public-facing for a larger organization, check with whoever handles compliance rather than assuming auto-captions satisfy it.

What about subtitles in more than one language?

If your audience watches in a language other than the one the video was recorded in, that's a separate job from captioning, not an extra setting on the same one. Scribe v2 transcribes in 90+ languages, so it can produce an accurate same-language transcript for a video recorded in Arabic, Hindi or dozens of other languages just as well as English, but it doesn't translate that transcript into a different language on its own. For translated subtitles, either run the original-language transcript through a separate translation step before formatting it as SRT, or use a dedicated dubbing/translation tool built for that specific job if you want translated audio as well as translated text. Treat "accurate captions in the original language" and "captions translated into a second language" as two different problems with two different tools, even though they look similar on the surface.

Step by step

1. If the video is going on YouTube, try its free auto-captions first and watch the result with the volume off.
2. List out every proper noun, product name and number that appears in the audio before you check the captions, so you know exactly what to look for.
3. Fix what's wrong in YouTube's own caption editor if you're using auto-captions, or in whatever tool produced your SRT file otherwise.
4. If auto-captions aren't accurate enough or your video isn't on YouTube, run the audio through Scribe v2 (or a tool built on it) to get a timestamped transcript, then format or export it as an SRT file.
5. Watch the full video with captions on, out loud, before you call it finished. Reading the transcript alone misses timing issues that watching catches.

When to skip the tool, and just check by hand

For a handful of short internal videos, manually typing captions or lightly editing free auto-captions is faster than setting up an API pipeline. Reach for Scribe v2 once you're producing videos regularly enough that the per-video manual effort adds up, or once free auto-captions consistently aren't accurate enough for your specific content (heavy jargon, uncommon names, multiple languages). It's also fine to skip the API entirely and just pay for a captioning service's per-minute rate if you only need this a handful of times a year; the setup effort of calling an API directly rarely pays off at that volume, even though the per-minute cost looks higher on paper.

How I'd approach this

If you're publishing video regularly, I've built small pipelines that take a finished video, pull the audio, transcribe it, generate the SRT file, and drop it next to the video automatically, so captioning stops being a manual step anyone has to remember. That's worth it once captioning is a recurring part of your workflow rather than an occasional task. For occasional videos, the manual process above is genuinely fine, and I'll tell you that directly rather than sell you a pipeline you don't need yet. Where this gets more involved is multi-language content: if a business is translating the same core training or marketing videos into two or three languages regularly, I'd build that as one pipeline, transcribe once, translate into each target language, format each into its own SRT, rather than treating every language as a separate manual job every time the source video changes.