Supported languages

Languages CaptionFX can caption and translate into.

CaptionFX transcribes eight spoken languages plus auto-detect, and translates captions into 12 Indian and 15 global target languages. Quality differs by language, and this page says where.

Short answer

CaptionFX transcribes Hindi, English, Korean, Spanish, French, German, Japanese and Chinese, with auto-detect, and translates captions into 12 Indian languages and 15 global ones. Hindi is its best-supported Indic language: clean Hindi reaches about 85-90% word accuracy and clean English about 95%.

Last updated

Try it freeSee features

Spoken languages

Hindi, English, Korean, Spanish, French, German, Japanese and Chinese, plus auto-detect for clearly spoken single-language audio.

Indian translation targets

Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese and Urdu.

Global translation targets

Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Turkish, Indonesian, Vietnamese, Japanese, Korean, Chinese and Arabic.

Which languages can CaptionFX transcribe?

The spoken languages you can choose in the studio today are Hindi, English, Korean, Spanish, French, German, Japanese and Chinese. Auto-detect works well for clearly spoken, single-language audio; for heavily code-mixed clips, naming the language yourself is more reliable.

Which languages can captions be translated into?

CaptionFX translates into 12 Indian languages — Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese and Urdu — and 15 global languages: Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Turkish, Indonesian, Vietnamese, Japanese, Korean, Chinese and Arabic. English is available as a translation target too.

Translation works on whole caption segments, never word by word, and the translated lines keep their timings. Indian and global targets use different translation systems, so quality is not identical across the list.

Does CaptionFX support Hinglish?

Yes, as romanized output. Hindi speech is recognised as Devanagari and then transliterated into Latin script, which is why a given word is spelled the same way each time. Code-mixed audio is difficult for every captioning tool, so treat Hinglish captions as a draft and review them before you export.

How accurate is each language?

Clean English audio reaches roughly 95% of words correct and clean Hindi roughly 85-90%. CaptionFX does not publish a single accuracy figure for the other languages, because it depends heavily on the audio and the language. Background music, several speakers and heavy accents all lower it, so review the captions before posting.

How it works

  1. Upload your videoAny common video format.
  2. Pick the spoken languageChoose it from the list, or leave auto-detect on for clearly spoken single-language audio.
  3. Pick the caption outputKeep the original language, romanize Hindi as Hinglish, or translate into another language.
  4. Review the captionsCheck names, brands and loanwords first; that is where errors concentrate.
  5. ExportDownload an MP4 with the captions burned in.

Questions people ask

Which languages can CaptionFX transcribe?

Hindi, English, Korean, Spanish, French, German, Japanese and Chinese, plus auto-detect. These are the spoken languages selectable in the studio today.

Does CaptionFX support Tamil, Telugu, Bengali or Marathi?

As translation targets, yes: you can translate captions into Tamil, Telugu, Bengali, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese and Urdu. Transcribing speech that is spoken in those languages is not selectable in the studio at the moment.

Can I translate English captions into Hindi?

Yes. Choose Hindi as the output language and CaptionFX translates the captions a segment at a time, which you can then edit over the video. Translated captions are 15 credits per video minute.

Does CaptionFX support Hinglish?

Yes. Hindi speech is transcribed into Devanagari and romanized into Latin script. The output is a draft that should be reviewed, because code-mixed speech is hard for every tool.

Is auto-detect reliable?

For clearly spoken, single-language audio, yes. For clips that mix languages, naming the spoken language yourself gives more reliable results.