Project · Python · 2026
YouTube Polyglot
A small web app that turns a YouTube video into text, then translates it into one of ten Indian languages.
How it works
Paste a link
A YouTube URL, plus the language the video is in and the language you want to read it in.
Pull the audio
yt-dlp downloads the best audio track, and FFmpeg converts it to WAV.
Transcribe
Sarvam AI's saaras:v3 speech-to-text model writes out what was said.
Translate
Sarvam's sarvam-translate:v1 model translates the transcript. If both languages match, this step is skipped.
Languages
Videos can be in English or any of these ten languages, and transcripts can be translated into any of the ten.
- Hindi
- Bengali
- Gujarati
- Kannada
- Malayalam
- Marathi
- Odia
- Punjabi
- Tamil
- Telugu
Details
It runs on Streamlit, in a single Python file. You bring your own Sarvam API key, which is entered in a password field and sent only to Sarvam. The temporary audio file is deleted after every run.
- Python
- Streamlit
- Sarvam AI
- yt-dlp
- FFmpeg