Browser-based subtitle generators, compared
Fifteen tools claim to make subtitles in your browser. They differ in one thing that matters: where the audio actually goes. Here is what each one does, checked in September 2026.
Advanced options
This tool is built and verified for English. Other languages run on the same small model and are noticeably less accurate — expect to correct a lot of the output. Chinese comes out in Traditional characters.
Your subtitles
Click any line to jump the player. Edit text directly — changes apply to the download.
How to read this comparison
Every tool below runs the speech model in the browser or on your own machine. None of them uploads your media as the default path. The difference is in what they charge for, what they let you export, and how much they make you download first.
Disclosure: this page is published by OfflineSubs, which is one of the tools listed. We have tried to describe the others accurately and to say plainly where we are weaker. Check the pricing pages yourself before you commit to anything — plans change.
The comparison
| Tool | Runs where | Limits on the free tier | Exports | Notable |
|---|---|---|---|---|
| OfflineSubs (this site) | Browser, WASM | None — no sign-up, no quota, no length limit | SRT, VTT, TXT, ASS, SSA, LRC, SBV, SMI, TTML, CSV | ~75 MB model, 0.6× realtime measured, batch queue |
| AudioSRT | Browser, or cloud on paid plans | None — but exports carry a watermark | SRT, VTT, TXT (Premiere JSON and burned MP4 on paid) | Full waveform subtitle editor; Windows desktop app claims 20× |
| SubtitleKit | Browser | None; 2 GB per file | 10 formats incl. ASS, LRC, TTML | 22 interface languages, picks model size by device, converts Chinese to Simplified |
| TranscriptSnap | Browser | None, no watermark | SRT, VTT, TXT | 17 languages selectable, including Chinese, Japanese and Korean |
| Subwave | Browser | None | SRT, VTT, ASS + burned-in video | Open source (MIT); on-device translation |
| VidStudio | Browser, FFmpeg WASM | None; 2048 MB per file | SRT, VTT + burned-in | Full video editor around the tool |
| FreeLabTools | Browser | None; under 30 min recommended | SRT, VTT, TXT, JSON | Lets you pick Tiny / Base / Small explicitly |
| Zalt | Browser | None, no rate limits | SRT, VTT | Three model sizes (45 / 80 / 250 MB) |
| WavyVid | Browser, WASM + WebGPU | None, no per-minute cap | SRT, VTT + burned-in | Big surrounding toolkit (compress, denoise, cut silence) |
| EchoWave | Browser | Not stated | SRT, VTT, TXT, JSON | Editable cues |
| SimpleTool | Browser | Not stated | SRT, VTT | Minimal single-purpose page |
| Meetrix | Browser | Not stated | Subtitle export | Whisper running in the browser |
| Whisper Web (whisperweb.dev) | Browser on the free mode | Free mode is local; the paid mode uploads | SRT, VTT | Hybrid: local for free, cloud for speed |
| SnailText | Desktop app (Mac / Windows) | Free tier uses the compact models | SRT, VTT | No watermark, works fully offline, batch folder transcription |
| Whisper Web (whisperweb.app) | Browser | None | Text only — no subtitle export | The best-known one, but English-only models and no SRT |
Where browser-based tools genuinely win
- Nothing to meter. The compute happens on your CPU, so there is no per-minute cost to pass on. That is why almost every tool on this list is unlimited rather than capped.
- Nothing to hand over. Unreleased footage, client material and internal recordings never reach someone else's disk.
- No queue. Cloud tools make you wait behind other jobs (Video Transcriber AI allows 4 free transcriptions a day; TurboScribe 3 files of 30 minutes).
Where they lose — including us
- Speed. Browser inference is CPU-bound and single-threaded. We measure about 0.6× realtime — a 10-minute clip takes roughly 20 minutes. A desktop app with GPU access (AudioSRT's Windows build) claims 20×.
- First-use download. 45–466 MB depending on the model you pick. Ours is ~75 MB.
- Languages. The browser bundles are the smaller Whisper sizes. Accuracy on non-English audio is well below English. OfflineSubs currently targets English only.
- Long files. Audio is decoded into memory. Very long recordings can exhaust it — split them.
What to check before you pick one
- Open DevTools, go to the Network panel, drop a file in and run it. If you see requests carrying your audio, it is not local.
- Look for the watermark — several free tiers add one to the export.
- Check which formats you actually need. If you deliver to a broadcaster you need TTML; if you only need YouTube, SRT is enough.
FAQ
Are these tools really processing files locally?
Most of them, yes — that is the whole point of the category. But naming is unreliable: whisperweb.tech carries a "Whisper Web" name while running transcription through a cloud provider. Verify in the Network panel rather than trusting the name or the marketing copy.
Why are so many of them unlimited?
Because your CPU pays for the compute, not theirs. There is no per-minute cost to meter, so a quota would only exist to push you towards a paid plan.
Which one is fastest?
A desktop app with GPU acceleration, by a wide margin. Inside the browser everything is CPU-bound and single-threaded, so the differences are smaller than the model size you choose.
Do I need to install anything?
No. All the browser tools here download a model on first use and cache it. Only SnailText is an installed app.
How was this compared?
Each entry was checked against the tool's own page or pricing page in September 2026. Limits and prices change — verify before relying on them.