Vaani

Vaani is an AI dubbing platform that produces localized video without asking you to re-record a narrator. It takes a source video, clones the voice of the person speaking in it, translates the transcript, and renders a finished MP4 with the cloned voice mixed against the original music and effects. Because the voice model is built from the source audio itself, a few seconds of someone talking on camera is generally enough to clone them.

The translation layer is scene-aware rather than sentence-by-sentence: the model reads the visual context of each shot so idioms, cultural references and tone adapt to what is on screen. Vaani covers 13 Indic languages — Hindi, Tamil, Telugu, Kannada, Malayalam, Bengali, Marathi, Gujarati, Punjabi, Urdu, Odia, Sanskrit and Nepali — plus 27 global languages including Spanish, French, German, Japanese, Korean, Mandarin, Arabic, Portuguese and Italian. Transcription includes speaker diarization so multi-speaker footage keeps each voice distinct, and every line can be reviewed or edited before rendering.

Lip-sync is optional and targeted at talking-head footage. Face-aware slicing means only frames where a face is visible are billed, while B-roll and cutaways pass through untouched. Two engines are available: Normal lip-sync built on Replicate sync/lipsync-2 for cost-sensitive work, and Pro lip-sync built on sync.so sync-3 for sharper edges, profile angles and fast-cut robustness. The platform serves creators, broadcasters, OTT platforms, EdTech companies, news publishers and studios, and offers a REST API plus webhooks from the Studio plan upward, with enterprise deployments supporting on-premise or private cloud, custom voice libraries, SLAs, SSO and audit logging.