How to subtitle your videos in several languages with AI

TLDR: Subtitling in several languages is no longer a budget line, it is a process. Gladia transcribes by hour of audio with free welcome credits, Descript fixes what the machine misheard, and Submagic applies the style and translates into whatever languages you want. The step nobody should skip is the human review: machine translation of proper nouns and numbers is where the result falls apart.

This guide is for anyone publishing video with speech who wants it understood: without sound, in another language or by people who are hard of hearing. Subtitles are not an accessibility extra, they are the condition for video to work on social, where most consumption happens in silence.

The underlying decision, and the one that orders everything else, is whether your volume justifies building a process or whether subtitling piece by piece is enough. This guide covers both routes.

1. Decide whether you need a tool or a pipeline

The boundary sits at hours per month. If you publish under five hours of video, subtitling with an interface tool is cheaper in time and money: upload, proof, export. If you publish dozens of hours (media, training, back catalog), building an automatic process that transcribes on upload changes the order of magnitude of the cost.

The second question is languages. Translating subtitles is cheap on either route, but it only makes sense if you will maintain those markets: publishing in five languages and answering comments in one is a worse signal than publishing in two and serving both.

With those two answers, the rest of the pipeline decides itself.

2. Transcribe the material with Gladia

For volume, Gladia is the economical piece: a transcription API billed per hour of audio processed rather than a monthly fee, with speaker separation and timestamps. Free welcome credits cover dozens of hours, so you can measure the real cost of your archive before committing.

The condition of use is clear: it is an API, it needs someone to integrate it. In exchange, a month without publishing costs zero and a conference month with twenty sessions costs exactly what those sessions run. For a media outlet with an archive or a training platform, that difference against subscriptions is an order of magnitude.

If you have nobody to wire it up, skip this step: Descript’s transcription in the next step covers the low-volume case without writing code.

3. Proof and fix the transcript in Descript

This is the step separating a professional subtitle from one that exposes the shortcut. Descript shows the transcript next to the video, so proofing means reading and fixing where it grates. And it grates in the same places every time: proper nouns, brands, industry jargon, spoken numbers and any moment with two people talking at once.

It has a free plan with 60 minutes a month and paid from $16, and if your volume is low it doubles as the transcriber, saving you the previous step. Its added advantage is that you can fix the video while proofing: removing filler words and silences improves both the subtitle and the piece.

Practical rule: always review the first minute with a magnifying glass. That is where someone decides to keep watching, and an error there weighs more than ten at the end.

4. Generate and translate subtitles with Submagic

With correct text in hand, Submagic adds the visual layer: animated subtitles using templates in the format that performs on social, plus translation into other languages. From $12 a month annually, with a watermarked free tier to test the style.

On translation, a warning that covers this whole guide: machines translate ordinary sentences well and fail at exactly what shows most, proper nouns, local expressions and wordplay. If a market genuinely matters to you, have someone who speaks the language review the result before publishing. If it does not matter enough for that, perhaps you should not publish in that language.

Subtitled and translated video fits the rest of the content pipeline: for slicing it into clips there is the clips from long video guide, and for full voice dubbing, dubbing videos into another language. The whole sector lives in AI for content and media.

Common mistakes

Publishing automatic subtitles unproofed. This is the mistake that defines careless content. Transcription hits often and fails precisely on the words that matter most: the ones specific to your subject.

Translating into languages you will not serve. A video in German with unanswered German comments projects abandonment. Every published language is a commitment, not a checkbox.

Subtitling without thinking about the format. The subtitle of a long horizontal video and that of a 30-second vertical are nothing alike: one accompanies, the other is the visual lead.

Letting subtitles sit under the interface. In vertical, the bottom of the screen belongs to the social network’s controls. Subtitles go center, not bottom, or nobody reads them.

Frequently asked questions

How much does subtitling an hour of video cost?

It depends on the route. With hourly transcription the cost per hour of audio is cents, plus generous free welcome credits. With subscription tools, the fee starts at $12 to $16 a month with minute caps. For high volume, the first route wins clearly.

Is automatic transcription reliable?

Good with clean audio and a single voice, worse with strong accents, overlaps and technical vocabulary. Enough as a draft, not enough to publish unproofed.

Is translating subtitles worth it?

If you have or want an audience in that language, it is the cheapest way to reach it: same video, new market. On the condition that you review the translation, because proper noun errors are what expose the shortcut fastest.

Burned-in subtitles or a separate file?

For social networks, burned in: it guarantees they show and allows animated styling. For long-video platforms, a separate file, because viewers can pick a language and search engines read the text.