Jack SaundersBreaking Down Synthesia 2.0We take a look at the tech that's underpinning the impressive new launch.Jun 24Jun 24
Jack SaundersinTowards Data ScienceScale Is All You Need for Lip-Sync?Alibaba’s EMO and Microsoft’s VASA-1 are crazy good. Let’s break down how they work.Jun 7Jun 7
Jack SaundersThe Definitive Guide to Lip Sync CompaniesThere are now so many lip-sync/AI Avatar/ Talking Face Generation Companies. Who is who?May 251May 251
Jack SaundersinTowards Data ScienceGaussian Head Avatars: A SummaryThere has been a recent explosion of Gaussian Splatting papers and the avatar space is no exception. How do they work and are they going to…Dec 19, 2023Dec 19, 2023
Jack SaundersinTowards AIREAD Avatars: Realistic Emotion-controllable Audio Driven AvatarsAdding Emotional Control to Audio-Driven DeepfakesAug 25, 2023Aug 25, 2023
Jack SaundersinTowards AIDAE Talking: High Fidelity Speech-Driven Talking Face Generation with Diffusion AutoencoderDiffusion Models + Lots of Data = Practically Perfect Talking Head GenerationJul 14, 2023Jul 14, 2023
Jack SaundersinTowards AITowards Generating Ultra-High Resolution Talking-Face Videos with Lip-SynchronizationThe holy grail of deepfake models is the person-generic model. So far no person-generic model has been of high visualquality, this paper is.May 27, 2023May 27, 2023
Jack SaundersinTowards Data ScienceShould Deepfakes be Open-Sourced?The purpose of this article is to try and start a conversation about open-sourcing in deepfake research. I cover the pros and cons.May 25, 20232May 25, 20232
Jack SaundersPerson-Specific Deepfakes with 3D Morphable ModelsWhile person-generic models such as Wav2Lip are incredibly versatile and work out of the box with any face and any audio, they still lack…May 11, 2023May 11, 2023
Jack SaundersWav2Lip: Generalized Lip Sync ModelsWav2Lip is one of the most popular lip sync models. In this article we explore how it works and analyse its potential for improvment.Apr 27, 20231Apr 27, 20231