⚡️ Introducing DETECT-2B, the most advanced deepfake detection model


Building Synthetic Linus from Linus Tech Tips

Jul 8, 2020

On July 8th, Linus Tech Tips posted a video on his channel ShortCircuit, reviewing the CLEAVE keyboard from Truly Ergonomic. This entire video was produced without Linus, using synthetic audio and video technology.

We worked with the Linus Media Group (LMG) to take existing audio of Linus and created a synthetic voice that could be powered by text-to-speech. By using Resemble, LMG was able to remove key barriers when it comes to creating audio content – including no studio time, more flexibility on the script, changes in post-production etc.

Moreover, with gradients for emotion, and a new feature called CopyPaste, we were able to control Linus’ emotion to make him sound as natural as possible.

With Resemble’s web platform, we were able to quickly iterate on the script, and fine-tune the parts of the performance that the team wanted to tweak. More than the Emotion Gradients that we introduced last month, we also demonstrated how CopyPaste, a new approach to speech synthesis could take another audio clip and extract speech patterns to apply directly to the target speaker. This enabled the team to get really creative with the audio output.

By using CopyPaste, we could get multiple variants of speech, where the input to produce the variant was a combination of text and the voice over from James Strieb, writer at LMG. We can then take the recordings from James and control exactly how much of it we want to copy over to the target audio. This can range from 0% James – 100% Linus to 100% James – 0% Linus. The important part here is that in either scenario, Linus retains his voice identity.

From there, Linus’ team used DeepFaceLab to map the visuals with the voice. You can watch their entire process here:
To join our beta for CopyPaste, feel free to reach out to [email protected], or build your own free synthetic voice at resemble.ai.

More From This Category