How to Add Captions to Any Video for Free — Without Uploading It

By Free Auto Captions AI · June 2026 · 5 min read

Captions used to be a slog. You'd either pay someone to transcribe the audio, use a subscription tool like VEED or Kapwing, or spend hours doing it manually. The cheaper tools gave you watermarks. The good ones gave you a monthly bill.

That's shifted. OpenAI released Whisper in 2022 under an open-source licence, and since then the community has spent a lot of effort getting it to run directly in the browser. It works. You can now get close-to-Whisper accuracy, for free, with nothing uploaded to a server, in a regular browser tab.

This article covers how it works, where the limits are, and what to actually use.

Why captions matter more now

The case for captioning has shifted from accessibility (still important) to engagement. The data on this is pretty clear: a large portion of social media video is watched on mute, and captioned videos consistently outperform uncaptioned ones on every major platform. If you're producing video content without captions in 2026, you're leaving reach on the table.

The problem with most captioning tools is that they're subscription products built for teams with budgets, not individuals making content on their own. VEED starts at around €18/month for meaningful usage. Kapwing limits you to 10 minutes of auto-subtitling per month on the free plan and watermarks your exports. Even tools that describe themselves as "free" tend to be free only up to a point that becomes inconvenient quickly.

What browser-based Whisper actually is

Whisper is a speech recognition model trained by OpenAI on a large multilingual dataset. It's accurate enough to handle accents, background noise, and technical vocabulary better than most competing models. OpenAI released it under an open-source licence, which is why it has been widely adopted.

The breakthrough for browser-based use came from the Transformers.js library, which allows machine learning models to run in the browser via WebAssembly. This means the same Whisper model that runs on a server can now run in your browser tab, with no server involved at all.

The tradeoff is that the model needs to download once (~75 MB) before it works. After that first download it lives in your browser cache and loads instantly on every subsequent visit.

How to caption a video with Free Auto Captions AI — step by step

How it compares to the subscription tools

Accuracy expectations

Whisper base is good, not perfect. Clear English with minimal background noise comes out at something like 90%+ accuracy in my experience. It handles German, French, Spanish, and most major European languages well, and auto-detects the language without you having to set it.

It breaks down on fast speech, strong regional accents, overlapping voices, and bad audio. If your recording has a lot of room echo, expect to spend some time in the editor cleaning it up. That's not unique to browser-based tools — cloud services have the same problem, they just charge you while they fail.

For broadcast or legal transcription, use a human reviewer. For YouTube, social video, internal presentations, or accessibility on your own content — this is production-ready.