AI dubbing struggled with fast cuts: Synthesia reframes

Synthesia Dubbing 2.0 upgrades AI video translation with frame-by-frame lip-syncing for quick cuts and multi-voice scenes across over 130 languages.

Synthesia is overhauling its dubbing component with Dubbing 2.0, a version that the publisher believes delivers a first render that is directly publishable for most videos, from a source language to over 130 others, without heavy editing or going through a localization agency.

The most visible improvement is in lip-syncing. Where the previous model struggled with quick cuts, camera angle changes, and multi-voice scenes, the new version tracks micro-movements of the mouth to stay aligned frame-by-frame, even between two languages with vastly different articulations like English and Japanese. Meanwhile, the audio engine gains in naturalness, with a pacing closer to real speech and better-rendered accents, while the emotional range expands enough so that a product pitch and a compliance note no longer sound with the same tone.

The translation now respects the original timing and length, which reduces the need to extend or manually rework segments, and glossary management is finally applied to dubbing to keep product names and recurring formulations consistent from one language to another. Editing also shifts its logic: proofreading transcriptions and translations no longer consumes credits with each pass, and a faulty segment regenerates on its own, without restarting the entire video.

The entire suite is open to all accounts, with enterprise plans unlocking frictionless editing and unlimited dubbing.