AI Auto Captions Generator
Turn speech in your Reel, Short or YouTube video into styled, word-by-word captions, right in your browser. On-device AI · your video is never uploaded
1 · Your video
Tip: drag the captions up or down on the preview to place them.
2 · Auto-transcribe
Choose a video, pick the language, press Generate captions.
3 · Edit captions
4 · Style
5 · Export
SRT works in CapCut, VN, Premiere Pro, DaVinci Resolve, YouTube Studio and Facebook.
AI makes mistakes, especially with names, numbers and mixed Hindi-English. Play it through and fix words before you publish.
How it works
How to use the ai auto captions
Choose a video and the language. The open-source Whisper speech model runs inside your browser (WebGPU when your phone or laptop supports it, otherwise the CPU), so the video itself never leaves your device. The first run downloads the model once, about 77 MB for Base or about 250 MB for Small; after that it loads from your browser's cache. Hinglish mode transcribes Hindi and then writes it in English letters, the way most Reels captions are typed.
Captions come out with word-level timing, grouped into short lines (3 words is the Reels sweet spot; change it from 1 to 8). Fix any word, nudge start and end times, split or merge lines, or find and replace a name everywhere. Pick a style such as Bold Reels with a yellow word highlight, Clean Outline, Box, Hindi Devanagari, Neon or a minimal lower-third, then drag the captions to where you want them on the preview.
Download an SRT or VTT file for CapCut, VN, Premiere Pro or YouTube Studio, a plain text transcript, or export the video with the captions burned in. Burned-in export plays the video once in real time, so a 60-second Reel takes about a minute; keep the tab open. Long videos take longer to transcribe on phones, so try a short clip first and always check names, numbers and technical words before you publish.
- Choose your video or audio file.
- Pick Hindi, Hinglish, English or Auto-detect, and the Base or Small model.
- Press Generate captions and wait for the progress bar (the model downloads only the first time).
- Edit any wrong words, timings or line breaks while the preview plays.
- Choose a style, drag the captions into place, then download SRT/VTT/TXT or export the captioned video.
FAQ
Questions about this tool
Does it support Hindi and Hinglish captions?
Yes. Choose Hindi for Devanagari captions or Hinglish to get Hindi speech written in English letters (for example “aaj hum baat karenge”). The Whisper Small model is noticeably more accurate for Hindi than Base, but it is a bigger one-time download. Hinglish spelling comes from a built-in converter, so give names and brand words a quick check.
Is the auto caption generator really free?
Yes. There is no sign-up, watermark or minute limit, because the AI runs on your own device instead of a paid server. The only cost is the one-time model download (about 77 MB for Base) and your device's processing time.
Is my video uploaded anywhere?
No. Your video and audio stay in your browser. Only the open-source Whisper model files and the transformers.js library are downloaded from public CDNs (Hugging Face and jsDelivr); nothing you record or upload is sent to this site or anyone else.
Which export formats can I download?
SRT and WebVTT subtitle files, a plain TXT transcript, and a video with the captions burned in. The video saves as MP4 when your browser can record MP4 (recent Chrome on Android and desktop) and as WebM otherwise; SRT imports into CapCut, VN, Premiere Pro, DaVinci Resolve and YouTube Studio.
Keep going
More creator tools
Related services
Hisar → India → Worldwide · Hindi & English
Need this done professionally? WhatsApp Natwar
Free tools get you started. For finished, on-brand work, send your brief and get a clear quote before anything starts.