Auto Captions for Clearer, Easier-to-Follow Videos

How to auto generate captions for a video

Generate captions from the speech in a video, then choose an export suited to your use. Burned captions become part of the video picture; SRT or VTT files remain separate subtitle tracks. For burned captions, adjust the text size, colors, background and position. Review names and timing before publishing, especially when speech overlaps or background noise is strong.

Cost: Free within the available allowance. Usage limits apply; see the guide below.

Add auto captions to a video in five steps

Upload one video
Click Upload a video to add captions or drop the file onto the preview. MP4, MOV, MKV, WebM, AVI, M4V, MPEG and OGV files up to 200 MB work, as long as the video has a sound track. A video you already used in another tool is listed under Uploads.
Set the spoken language
In the Captions card, Language starts at Auto, which lets the speech model detect the spoken language. You can also choose a specific language code. Captions are written in the language that is spoken; this tool does not translate them.
Choose the output
Under Output, Burn in returns an MP4 with the captions drawn into the picture. SRT and VTT return a subtitle file with timestamps and leave your video as it is.
Adjust the style
In the Style card, Words/line sets the maximum number of words per caption for every output. For Burn in, also set Font size, the Text color and the position with Top, Center or Bottom. The sample line on the preview shows the color and position.
Generate, check, download
Click Generate captions and wait for the job to finish. A burned-in result plays in the preview, so you can check names, technical terms and timing before you click Download. The captioned MP4 is also added to Uploads for the other video tools.

Burn in, SRT or VTT?

All three outputs come from the same transcription. They differ in where the text ends up and what you can change later.

Burn in: captions inside the picture

Viewers see the captions in every player and app, including feeds that start videos without sound, and nobody can switch them off. The text can no longer be edited or translated, and the whole video is re-encoded.

SRT: a separate subtitle text file

A plain text file with numbered captions and time ranges such as 00:00:04,250 --> 00:00:06,900. Check whether your video editor, player or upload form accepts SRT. Viewers can turn the separate track on or off when the player supports it.

VTT: subtitles for web players

WebVTT is the format HTML5 video players load through the track element. It holds the same captions as the SRT file, with a WEBVTT header and dots instead of commas in the timestamps.

Caption settings explained

Language and Words/line affect every output. The other style settings only matter for Burn in.

Language: Auto is the default. You can also choose en, de, es, fr, it, pt, ja or zh (English, German, Spanish, French, Italian, Portuguese, Japanese, Chinese). A code tells the model which language to expect. Use Auto for other languages or when you are unsure.
Output: Burn in (default), SRT or VTT; the comparison above shows when to use which.
Font size: 10 to 72, default 20. The number is not a pixel value: text scales with the height of your video, so the same setting gives the same proportions at any resolution. See the sizing example below.
Words/line: 2 to 16, default 8. The maximum number of words in one caption. Longer captions are split into several, and their time is divided by word count. Japanese and Chinese don't separate words with spaces, so for them the speech model sizes captions from this value and caption length varies more.
Text: The caption text color, white (#ffffff) by default. The burned-in video uses exactly this color.
Background and Background opacity: Set the color of the box behind the text (default #000000) and its opacity (0–100 %, default 70 %). Lower opacity lets more of the picture show through the box; the burned-in video uses exactly these values.
Top, Center, Bottom: Places the captions centered horizontally at the top edge, in the middle or at the bottom edge of the frame. Bottom is the default. Choose Top when the lower part of the picture already shows text such as name titles or prices.

Font size on landscape and vertical videos

Font size is measured against the frame height, not in pixels. The default 20 gives a text line about 7 % of the frame height: roughly 50 px on a 1280 × 720 px video, 75 px on 1920 × 1080 px and 133 px on a vertical 1080 × 1920 px clip.

That works well for landscape video. On the vertical clip, the default 8 words per caption at size 20 wrap onto three lines and cover much of the lower third. As a rule of thumb, vertical videos read better with a font size of 12 to 14 and 4 or 5 words per caption.

The sample line on the preview scales with your video, so you can judge size and line breaks before you click Generate captions.

What happens to your video during captioning

FFmpeg first extracts the sound track as a 16 kHz mono WAV file. Only this audio is sent to Google's Gemini API, which returns the spoken words with start and end times. The picture itself is not sent anywhere.

The timestamps are the model's estimate, not a frame-exact alignment. The tool keeps every caption within the video's length, shortens captions that would overlap and splits long ones by Words/line. Expect occasional wrong names, missed words under loud music and captions that start slightly early or late.

For Burn in, FFmpeg draws the captions onto every frame and encodes a new MP4 with H.264 video at CRF 22 and AAC audio at 192 kbps. Resolution and frame rate stay the same, but the re-encode can lose some fine detail. The output uses 8-bit H.264, including for 10-bit HDR sources; check its colour and brightness against your source, because the result depends on the source's HDR characteristics.

Formats, limits and file handling

Input formats: MP4, MOV, MKV, WebM, AVI, M4V, MPEG/MPG, OGV with a sound track
Output formats: Burn in: MP4 (H.264 video, AAC audio). SRT or VTT: subtitle text file
File names: For an upload named interview.mov: interview_captioned.mp4, interview.srt or interview.vtt
Maximum file size: 200 MB per video
Files per job: 1
Processing: On our servers with FFmpeg. A job stops after 15 minutes, including transcription and burn-in.
Speech recognition: The extracted audio is sent to Google's Gemini API. The picture is not sent.
Watermark: None
Storage: Uploads and results are deleted automatically after 48 hours for guests and after 30 days with a free account. Paid plans can keep files longer.

Questions about auto captions

Can I use the auto caption generator for free?

Yes. Captioning costs no credits, works without an account, and the burned-in video has no watermark. Daily and hourly usage limits apply; they depend on whether you are signed in and on your plan.

Can the captions be translated into another language?

Not in this tool: captions are always written in the language that is spoken. Video to Subtitles creates SRT and VTT files, can translate them, lets you correct text and timing, and can then burn them into the video.

Why are some captions wrong or slightly out of sync?

Text and timing come from a speech model. Background music, people talking over each other, strong accents and uncommon names cause most mistakes, and a Language setting that doesn't match the speech makes them worse. To fix single lines, export SRT and edit it in any text editor: each caption is a numbered block with a time range and its text.

Is my whole video sent to an AI service?

No. Only the extracted sound track goes to Google's Gemini API for transcription; the picture stays on our server. Your upload and the result are deleted automatically after 48 hours as a guest or 30 days with a free account.

Why did caption generation fail?

The usual causes are a video without a sound track and audio without intelligible speech, such as music only; in that case the tool reports that no speech was detected. Long videos can also fail, because the job stops after 15 minutes. Trim the video to the part you need with Cut Video and try again.

Why is the captioned MP4 a different size than my original?

Burning in re-encodes the video and audio. The result can be smaller or larger, depending on the source content, encoding and bitrate. Compare the actual file sizes. Compress Video lets you adjust the result's compression.