Douglas McGinness’ Post

Friday animation test is up! 🚀 Today I'm diving into Live Portrait video2video, using Viggle as an animation style transfer and Epic Games Metahuman to drive the facial animation. Let me know your thoughts as always, and if anyone has used the AnimateDiff workflow that was posted about earlier today let me know as well... #generativeai #ai #animation #production #liveportrait #anime

Tomorrow, and tomorrow, and tomorrow... as performed by AI. Created with the help of Midjourney, Live Portrait, Epic Games Metahumans, and ElevenLabs. What are you making with LivePortrait? #generativeai #ai #filmmaking #production #animation #comfyui #stablediffusion

Happy Friday! Today's test is with the brand new LivePortrait ComfyUI node, which offers facial animation of a still image from a reference video - and it doesn't disappoint. I brought in an old Metahuman render I created awhile back as the reference video, to see how it would perform, and it seems to work relatively well. Most of these models are trained on live-action footage, so trying to bring in animation references is sometimes challenging, but in general I think this is a really interesting workflow for adding emotion to animation. ComfyUI link here: https://lnkd.in/eENK49Jw Anyone else getting good results with it so far? #animation #generativeai #ai #comfyui #stablediffusion #anime #openai

Today's eerie, moody test takes me back home to Kansas, where I grew up :) With tools like Runway's Gen-3 and Luma AI's Dream Machine its getting scarily easy to bring a concept to life. While Gen-3 is not as easy to control as Dream Machine, the quality of the results is really remarkable. While some of this story was shaped by the results in that parts of the edit was formed around key shots that Runway/Luma gave me, I'm happy to say that the majority of what was in my head made it onto the screen, and to me that's really impressive. What are your thoughts on Gen-3? #generativeai #ai #runway #gen3 #luma #dreammachine #production #animation

Today I'm testing Hedra, which offers lip-synced animation from a still image and an audio file. As always I'm most interested in how it works with animation, and I have to say it doesn't disappoint. I've brought in some voice acting from ElevenLabs, as well as actress Katherine de Leiros, and everything else is created from a mix of Midjourney, Cuebric, ToonCrafter, and Topaz Labs for upscaling. Getting close to being able to have a finished animation short here, which I'm very excited about. What's your experience with Hedra? #anime #animation #ai #generativeai #hedra #tooncrafter #production

If you can get past the high queue times 😅 what Luma AI Dream Machine does exceptionally well is combine 3D spatial data with AI interpretation - meaning it does an excellent job with consistency and logic. These are just some initial tests but its clear that the team is really on to something here. I'm not sure it quite kills OpenAI's Sora, but its definitely a tremendous step above Runway and Pika so far. As always I wanted to test it with animation as well as live action, and while the results are definitely hit and miss its a very interesting entry into the text/image to video race. What are your impressions with using it, and have you found any interesting tests yet? #lumai #generativeai #ai #videoproduction #animation #openai #sora #dreammachine

Exciting opportunity for AI filmmakers, run by some really talented people. Who's joining in? #filmfestival #ai #generativeai #competition