The actor’s performance in the close-up is incredible. But when you cut back to the wide, it doesn't match. The edit feels jarring, and it’s breaking the scene.
Your options are limited: either use a take you like less, or try to match the two shots as smoothly as possible. Neither is great.
Performance transfer – when you copy an actor’s facial performance from one shot and paste it onto another – is a core use case of visual dubbing.
When applied correctly – with the actor’s consent – the technique can be an extremely powerful way to solve problems in the edit.
As a result, your edit will have the shots, performances and impact you want.
This blog post explains:
Performance transfer is a key technique of visual dubbing and AI-assistive editing.
The technique is new, and the concept is simple: you take the performance from one take and map it onto the shot from another take.
Performance transfer stands apart as a DeepEditor use case because it’s the only one that requires a full audiovisual file as driving data. For other workflows, like visual ADR, audio alone is enough.
The actor's expressions, delivery, timing, and emotional intensity – you can move all of it to the shot you actually want to use.
Performance transfer shines in a few specific situations:
Simply upload the angle or setup you want to use as your source media, and then upload the take with the desired performance as your driving data.
Upload your new audio recording and original shot together in DeepEditor – either in the web app or via the Avid plugin.
DeepEditor will copy and paste the performance you want onto the shot you want, preserving all the facial movements.
Workflow TipIn DeepEditor, there are two types of vub output: draft vubs and final vubs. We recommend you use draft vubs for creative iteration and problem solving in the edit – they are affordably priced to let you use them in this way. They work in HD, making them perfect for offline editing. You can even remove the watermark for non-commercial purposes, like showing your director and producers your edit. Once everyone’s happy to commit, you can upgrade your vub to a final quality output (up to 8K, 16-bit and with lossless color), so the online edit has a truly flawless result. |
Because DeepEditor is assistive AI, not generative, every output from DeepEditor is fully controllable after creation. If needed, you have the option to use the DeepEditor refinement tool to make tweaks to the vub.
It’s also critical you get consent from the actor once your vub is created via the Artistic Rights Treasury (A.R.T.) to be compliant with guilds and unions.
Insert your DeepEditor vub into your timeline in Avid or other non-linear editor, and continue your edit!
Using performance transfer on a recent series, director Rajeev Dassani said, “I knew that in certain cases, if I didn't get the exact performance I needed in a wide shot, I can now move the close up shot to the wide shot."
That means every micro-expression, every subtle moment, every piece of emotional nuance comes from a real human performance, not a reconstruction.
For the first time in live-action, you can shape performance with the freedom of animation – while DeepEditor ensures the process is secure, transparent, and grounded in full actor consent.
DeepEditor offers several powerful use cases for film and TV production, each designed to solve specific post-production challenges:
Performance transfer addresses the age-old problem of choosing between technical quality and acting quality. But the underlying technology – frame-by-frame facial performance control – opens up entirely new ways to work.