XR Studios

I have been talking to IP holders, as well as investors about crafting a new vision for #cinema. Today, the first studio domino to fall, is Lionsgate, and you can bet that others will follow in order to compete. Soon, hyper-personalized series and movies, complete with ads and product placements, will be served to you, on demand. Some of them will even include you, your friends, and/or family, as you upload clips of your voices, videos, scans, etc., so the Generative Engine can insert you to be the leads, or supporting characters, of the programs you want to watch. In some instances, it might be only fully synthetic content. For studios with big catalogs, there is a lot of money to be made in letting people re-re-re-re-watch #StarWars with themselves as Han, Leia, or even Luke. Maybe *you* want to be Frodo to take the ring to Mordor, but this time, via an alternate route? This time, you’ll encounter new characters, new monsters, and depending on the pipeline, real-time interaction with fellow audience members who are also participants, on the other side of the world. Or, perhaps your kid wants to be Dorothy today in “The Wonderful Wizard of Oz.” Alternatively, perhaps a variant of the film opens up as an open-world in which you can participate, like #Cyberpunk2077. It’s a #marketing team’s dream come true, and a nightmare for everyone in legal. ?? Make no mistake that this is an end, as well as a new beginning. The revolution will be generated, not televised. We’re still open for discussions on various aspects of our creative and technical approaches. For those who are interested, please message me directly on LinkedIn. https://lnkd.in/gYzMUJF7 _ #lionsgate #xr #filmproduction #postproduction #vfx #streaming #ai #genai #generativevideo

#Arcane is one of those rare gems that is not only an #adaptation, but a masterful technical accomplishment, and a singular aesthetic achievement that elevates it into a masterpiece. Season 2 is finally coming to #Netflix. _ #adultanimation #vfx #videogame #leagueoflegends

After long months , I am so happy to share probably one of the biggest and most early project using mostly ComfyUI (in march) and other AI tool to generate a full animation movie. this project was a Real team project made with DOGSTUDIO/DEPT? for The Coca-Cola Company . second part IS making off ! Hope you will like it !

Elevating AI-Generated Video Quality with VEnhancer! This is incredibly powerful, generative video people, take a look! ?? Say hello to #VEnhancer, a new framework designed to improve the quality of AI-generated videos by enhancing both #spatial and #temporal resolution. Unlike traditional methods, VEnhancer improves existing #TextToVideo results by refining visual details and creating smoother motion— effectively eliminating artifacts and flickering! This is so much better than attempting to deflicker in #compositing, or in having to create start and end frames that are (not) super precise with #GenAI (which can be challenging), or without involving another suite of skills, or staff, with 3D. Built on a pretrained video diffusion model, VEnhancer uses a specialized #video #ControlNet to condition low-resolution, low-frame-rate videos, enabling flexible #upscaling in both space and time. Its innovative approach leverages #spacetime data augmentation and video-aware conditioning, allowing for stable, end-to-end training. The results speak for themselves: VEnhancer outperforms current state-of-the-art methods for video super-resolution, enabling tools like VideoCrafter-2 to achieve top rankings in #videogeneration benchmarks. This breakthrough makes VEnhancer an essential tool for anyone looking to push the boundaries of AI-generated content. For those seeking to improve the quality and fluidity of AI-generated videos, VEnhancer offers a powerful, flexible solution. #Github code download: https://lnkd.in/genb_qsW _ #VideoEnhancement #GenerativeAI #VideoEditing #CreativeTech #GenAI #GenAIVideo #AIvideo

Redefining Volumetric Video: Introducing "Robust Dual Gaussian Splatting" Volumetric video has the potential to reshape the future of digital experiences, offering fully immersive content that allows users to navigate freely within virtual environments. However, its adoption has been slowed by the challenges of current techniques—primarily the need for painstaking manual cleanup to stabilize complex 3D mesh sequences. The new “Dual Gaussian Splatting” method tackles these challenges head-on, providing a fresh approach to capturing and rendering human-centric volumetric videos. Utilizing 81 calibrated industrial cameras, this technique captures highly dynamic human performances with remarkable detail. At the core of this method is a dual Gaussian representation, which compresses motion and appearance data separately, allowing for efficient storage and real-time high-fidelity rendering. The results are impressive: up to a 120X compression ratio without sacrificing quality (!!!), enabling seamless playback of complex scenes. By avoiding the heavy manual interventions required in existing workflows, this method streamlines the process, significantly lowering the barrier to entry for creatives and developers. Additionally, the technique integrates smoothly with VR platforms and CG engines, enhancing both playback and editing in various immersive scenarios. From #interactive #storytelling to advanced gaming and beyond, “Dual Gaussian Splatting” could be the key to unlocking the full potential of volumetric video, making immersive experiences more accessible and engaging than ever before. As we continue to blur the lines between the digital and real worlds, methods like these offer a glimpse into the future of how we create, share, and experience content. There is so much more that can and needs to be discussed about this technique and the medium, but there is only so much room on LinkedIn to capture that. If you are interested in volumetric media for your production, which could also mean volumetric content for #virtualproduction, I’m available for consultations. "Robust Dual Gaussian Splatting for Immersive Human-centric Volumetric Videos", Project Page: https://lnkd.in/ga32H5qf _ #volumetricvideo #dualgaussiansplatting #gaussiansplatting #3DGS #4DGS #volumetric #XR #metaverse #vr #ar #mixedreality #virtual #immersive #production #postproduction #animation #vfx #streaming #videogames

98.5% like ratio! Thank you for your super positive response to my latest YouTube tutorial! I’m glad Blender is bringing value to your AI art workflows. If you’ve missed it, check out my latest workflow - how to create Depth ControlNets for Stable Diffusion. Works in any UI! Link to the vid is in the comments - make sure you sort by “Most Recent” so LinkedIn doesn’t hide it from you. #generativeAI #stablediffusion #controlnet

LLM-based open-world, video game engine!?? Not even two weeks ago, I shared #Google #DeepMind's #GameNGen running #Doom from a specialized #LLM. It should come as not surprise that similar efforts are underway elsewhere, and that #neuralrendering taking us to the next levels of #media and #entertainment. Be sure to stay for the end of the trailer demo below, where they basically show off a take on the insanely complex #GhostOfTsushima from #Sony, as a product of this engine, and it's jaw-dropping. #GameGen-O: is an amazing #GenAI model that’s reshaping how we think about creating #openworld #videogames. This is the first #diffusiontransformer model designed specifically for generating entire game worlds from scratch, simulating everything from character interactions to dynamic environments and complex storylines. GameGen-O doesn’t just generate content; it brings a new level of #interactive control, allowing for #realtime #gameplay #simulation. Behind GameGen-O is the first-ever Open-World Video Game Dataset, OGameData, built from the ground up by collecting data from over a hundred next-gen open-world games (#3Dmodels still aren't dead, because these need to be trained, and next-gen image fidelity, #vfx, etc., still need to be made for training, but in time, it won't be as necessary as at present). This extensive dataset is curated using an innovative pipeline that sorts, scores, filters, and captions game data to create a rich foundation for the model. One has to wonder which games, and whether consent was provided, because that's a thorny problem. The research paper informs that #LightSpeedStudios, which is a part of #Tencent, participated in this effort. A lot of this training took place in #PROC, where #copyright effectively goes out the window. The training of GameGen-O happens in two key stages. First, it undergoes pretraining on OGameData, learning to generate and extend video content in an open-domain setting. In the next stage, we fine-tune the model with a specialized module, #InstructNet, to ensure it can generate future game sequences based on structural instructions. GameGen-O represents a major leap forward in using #generativeAI models for video game creation. It opens up new possibilities by combining #creative generation with interactive capabilities, challenging the limitations of traditional #rendering techniques. This is not where it stops. This is not "just" the future of gaming, you are also looking at the imminent future of hyper-personalized #cinema, very likely starting as #branching narrative content for #streaming platforms. If you would like to discuss the latter, my team has been exploring this for a while. Project Page: https://lnkd.in/gcaYYAnE Github (code will be released soon, so bookmark it: https://lnkd.in/gTnrEbc3 Please DM me directly via LinkedIn, I cannot reply to emails. Thank you.

Flux now in Realtime. available in Krea with hundreds of styles included. free for everyone.

Hey everyone! Thanks for all the encouraging feedback! It means the world to us! We are working hard to get a set of tools into your hands that give you unparalleled control and creative freedom. If you have reached out to us and we haven’t gotten back to you, please be patient. This is all a bit overwhelming at the moment, but we will adjust.. give us a little more time to sort things out, thanks! Cheers, zebrapunk & gizmo