Reshooting a video to fix one bad line — or to test a different hook — is slow and expensive. It doesn't scale, and it definitely doesn't work for teams running automated pipelines.
Magnific is solving that with VEED Lip Sync 2.0.
What the integration does
Magnific now runs on VEED's Lip Sync 2.0 API enabling developers and technical teams to swap the audio in any video — a new hook, a corrected line, a re-read, a translated track — and a result with natural lip sync. No reshoot. No per-subject training.
The model is zero-shot: feed it any face and any new audio, and it re-syncs the performance while carrying over the emotion and speaking style from the source audio. That's what makes it usable for dubbing and localisation at scale, not just one-off fixes.
How it works
- Submit your input — a video file and a separate audio file (new hook, fix, re-read, or translation)
- Lip Sync 2.0 detects the face — and re-renders the mouth and lower face region to match the new audio
- Get your output — a lip-synced video that preserves the emotion and delivery style of the input audio, returned via API in seconds
Built for forward-facing subjects up to medium close-up, with support for side-angle (MCU) framing.
Who it's for
- Product and engineering teams at media platforms — building automated pipelines to fix or replace audio without manual re-recording
- Marketing and social teams at scale — programmatically testing different hooks on the same take for A/B testing or repurposing
- Localisation teams — translating and dubbing content while keeping the original visual performance intact
- Developers building on top of Lip Sync 2.0 — creating in-house or client-facing tools for dialogue and VO updates
Why Magnific chose Lip Sync 2.0
Most lip sync tools either require per-subject training or aren't built with API-first, programmatic use in mind. VEED Lip Sync 2.0 swaps the audio and keeps the performance, matching lips every time with no training or fine-tuning required.
It also holds up where most lip sync models break: a hand or object blocking the face, low light, fast camera movement, footage shot from multiple angles. Extreme angles and close-up mouth detail are where the difference versus competing models is most obvious.
"Magnific and VEED share the vision of making video creation accessible to creatives across all levels of expertise. Magnific boasts one of the best generative media platforms in the world, serving everyone from solo creators to major studios. At VEED, we're honoured to be partnering with the Magnific team with the launch of our Lipsync model, to further reduce the barrier to creating great video" — [Josh Goldman, COO, VEED]
"At Magnific we want every creative, whatever their level of expertise, to produce video at studio quality. VEED shares that ambition, and Lip Sync 2.0 proves it: a model that keeps the performance and changes everything else. We're delighted to partner with the VEED team and put this in the hands of millions of creators." — [Omar Pera, CPO, Magnific]
Good to know
Lip Sync 2.0 currently supports videos up to 10 minutes and 4K resolution, with a 5GB file size limit. It works with one active speaker per scene, and doesn't yet support non-human subjects. Bilabial sounds (p, b, m) are an active area of improvement.
About VEED's Lip Sync 2.0 API
Lip Sync 2.0 is VEED's zero-shot video-to-video model, available as an API for developers and platforms building video and audio products. It's priced at $0.07 per second of video processed, regardless of resolution. VEED's APIs power AI video and audio features inside leading creative products used by millions of marketers, editors, and developers worldwide.
Explore: Lip Sync 2.0 API · All VEED APIs · Get API access
Already building with Lipsync 2.0? Get API access — no training required.