Symbaiosis

Symbaiosis › Auto captions

Auto captions, word by word, burned in

The speech model writes down what you say and times every word. Pick a style, position and highlight colour, fix any word it got wrong, and export the video with captions burned in, plus SRT and VTT files.

  • Runs on your device
  • No upload, no account
  • No watermark
Before: 31 s, wide
After: cut and captioned
Made with this editor from the sample clip (computer voice, nothing uploaded).

Drop a video to caption it

or drop a file here · MP4, MOV or WebM · it stays on your device

Three caption styles that read well on phones

  • Bold pop: heavy white letters with a thick black outline, and the word being spoken grows slightly and changes colour. The short-form look; readable over any background.
  • Boxed: white text on a dark rounded box. Calmer, and the most legible over busy footage.
  • Clean: white text with a soft shadow. For interviews and talks where captions should stay out of the way.

Choose one, three or six words on screen at a time. Fewer words means bigger, faster captions that suit vertical video; six suits wide YouTube videos. Captions break at the end of sentences and at pauses, so each one reads as a phrase.

Accurate enough to publish, after a quick check

The transcript appears as the model works. Read through it once: tap a wrong word, choose “Fix text”, type the right one. Names, brands and jargon are the usual misses. Corrections change the captions only, never the timing.

Captions as files

Along with the video, you get an SRT file (for YouTube, Facebook and most editors) and a VTT file (for websites), both timed to the edited video, so you can upload captions as a separate track instead of, or as well as, burning them in.

Where to put captions

On vertical video, the bottom fifth of the screen is covered by the app's buttons and the caption of the post itself, and the right edge holds like and share buttons. “Lower” puts captions at 70% of the height, above that zone; “Middle” is safest if your platform's buttons are large. Keep important text out of the top 10% too.

Questions

What languages can it caption?

English with a fast English-only model, and 18 more languages (Spanish, French, German, Portuguese, Italian, Dutch, Polish, Turkish, Russian, Ukrainian, Swedish, Indonesian, Vietnamese, Hindi, Japanese, Korean, Chinese, Arabic) with a larger multilingual model. Choose the language before you add the video.

Can I caption without cutting anything?

Yes. In the Cut tab choose “Keep all” for pauses and untick “Cut filler words”. The video stays as it was, with captions added.

Why are “um” and “uh” missing from my captions?

They're cut from the video by default, so they're not captioned either. If you keep them, they appear in the captions too.

Is the transcription done in the cloud?

No. The model (Whisper, by OpenAI) is downloaded to your browser once and runs on your device. Your audio is never uploaded.

Guides