Subtitles and captions make a meaningful difference for video content. Studies consistently show that a large portion of social media videos are watched without sound — viewers on commutes, in shared spaces, or simply scrolling silently miss all the spoken content in your video unless it is also displayed as text. Adding subtitles addresses that directly. They also make your content accessible to deaf and hard-of-hearing viewers, help non-native speakers follow along, and improve comprehension for technical content with jargon or complex terms.
The difference between subtitles, captions, and burned-in text
These terms are sometimes used interchangeably but they are technically different. Subtitles are a separate text track that can be turned on or off by the viewer — the video file itself is unchanged. Captions are similar but designed specifically for accessibility, including descriptions of non-speech audio like [applause] or [music]. Burned-in (or hard-coded) subtitles are text rendered directly into the video frames, which cannot be turned off but also cannot be missed or suppressed by a platform.
For social media, burned-in subtitles are usually the better choice because autoplay on platforms like Instagram, TikTok, and Twitter does not enable optional subtitle tracks. For YouTube, optional subtitle tracks (SRT files) are preferred because YouTube can also use them for search indexing.
Generating subtitles with AI
Transcribing audio manually is time-consuming. AI speech recognition has improved to the point where it can produce highly accurate transcriptions of clear speech in English and many other languages without human intervention.
ClipZap's subtitle generator uses Whisper, an open-source speech recognition model developed by OpenAI. The model runs locally in your browser — your video is not sent anywhere. The audio is extracted from your video and processed on your device using WebAssembly and the Transformers.js library.
You choose a model size when you start. The Tiny model is the fastest to download and process but is less accurate, especially with background noise, accents, or technical terms. The Base model is a good middle ground. The Small and Medium models are more accurate but take longer to process and require more RAM. For most speech in a relatively quiet environment, Base is sufficient.
Downloading as SRT vs burning into the video
Once the transcription is complete, you have two options.
The first option is to download an SRT file. This is a plain text file that contains the subtitle text along with timestamps indicating when each line should appear. You can upload an SRT to YouTube as a caption track, import it into a video editor like DaVinci Resolve or Premiere, or keep it alongside the video for use in media players that support subtitle tracks.
The second option is to burn the subtitles directly into the video. The tool overlays the text onto the video frames and re-encodes the result. The text is then part of the video itself, which works on every platform and player without any additional steps. The trade-off is that burned-in subtitles cannot be edited or disabled after the fact.
Editing the transcript before burning
After transcription, you should review the output before burning. AI transcription is accurate but not perfect — it can mishear words, especially names, technical terms, or words in noisy audio. The tool lets you edit the text in each subtitle segment before committing to the final output. This step is worth the few minutes it takes because errors in burned-in subtitles are permanent.
What formats and languages are supported
The subtitle generator accepts MP4, MOV, WEBM, and most common video formats. The Whisper model supports transcription in a wide range of languages including English, Spanish, French, German, Portuguese, Arabic, Japanese, Chinese, and many others. For non-English speech, the larger model sizes (Small, Medium) will produce better results than Tiny or Base.