Don Allen Stevenson III’s Post

🚀 Bringing my 2D character drawings to life with AI video! Here’s how I did it: 1. Drawing: Started with Procreate to sketch out my character. 2. Generate a life-like version of my character with Midjourney . In MJ, I used my drawing as a character reference and asked for photorealistic versions of the character. 3. Took the best-generated image into Photoshop for some fine-tuning 4. 3D Animation: Imported the final image into Luma, where I directed the animation using various prompts. The result? A highly detailed 3D render that looks basically cinema quality to use as modeling inspiring for me to build out manually. These tools are so helpful, not only great for streamlining my workflow but also helping me visualize the big picture before diving into modeling/animating. With image generation tools like Midjourney and AI at Meta , Video tools like Luma AI , Runway and OpenAI , the possibilities are endless. Can’t wait to see what comes next! 🤯 #AI #Animation #DigitalArt #Creativity #Procreate #Photoshop #MidJourney #LumaAI #Runway #OpenAI #Innovation

🚀 Unleashing the Future of Toon Shading: Meet Diffutoon! 🎨✨ Introducing Diffutoon: High-Resolution Editable Toon Shading via Diffusion Models 🕺, a revolutionary approach transforming photorealistic videos into stunning anime styles. Crafted by the brilliant minds of Zhongjie Duan, Chengyu Wang, Cen Chen, Weining Qian, and Jun Huang, this innovation is set to redefine the world of animation. Why Diffutoon? 🤔 ✅ High-Resolution & Detailed: Capable of rendering videos in exceptional detail and high resolution. ✅ Editable: Includes an editing branch to modify content based on user prompts. ✅ Consistency & Quality: Enhances temporal consistency and visual quality, tackling challenges that existing methods face. How Does It Work? 🧠 Diffutoon dissects the toon shading task into four key areas: 1. Stylization: Transforming video frames into anime style using advanced diffusion models. 2. Consistency Enhancement: Ensuring smooth transitions and consistent animation quality. 3. Structure Guidance: Maintaining structural integrity using outline information. 4. Colorization: Enhancing colors for a vivid and appealing visual experience. Try It Out! 🎉 Explore the capabilities of Diffutoon: • 🌐 Project Page: https://lnkd.in/g_9F9UDd • 📄 Research Paper: https://lnkd.in/gRJkPNww • 🧬 Source Code:https://lnkd.in/gN_kHbJJ • 🍊 Jupyter Notebook by camenduru: https://lnkd.in/gkckrcEU Huge thanks to the incredible team behind this innovation. Your work is paving the way for a new era in animation technology! 🙌 #AI #MachineLearning #Animation #ToonShading #DiffusionModels

Another day, another AI tool! This time, we’ve got something interesting: code, a paper, and a demo are all available. 👇 ToonCrafter uses live-action video data in a generative approach to address these issues. It includes a dual-reference 3D decoder to maintain fine details and a flexible sketch encoder for user control. Tests show ToonCrafter produces more natural animations, handling tricky parts better than existing methods. 🎨 In classic animation, there were three key roles: inbetweeners, inkers, and colorists. The animation director would draw key frames, capturing essential movements, while others would complete all the other steps. I was impressed by the quality of interpolation, which involves generating intermediate frames between the first and final ones through a trained latent space. This non-linear approach, trained on cartoons and videos, is impressive. 🤯 This method could significantly speed up cartoon production, especially for anime with minimal movement but expressive features. However, the model can make mistakes with background objects, causing erratic movements. Don't forget to check the comments for all the additional links! 🙌 #TechNews #AINews

I continue testing my new pipeline of making digital art by utilising AI-simulated sequences based on some of my previous 3D art experiments. This time, I took a freshly krea.ai generated animation as a texture to drive geometry displacement in Maxon Cinema 4D and rendered everything with Redshift. It helped me achieve a great level of fluidity in motion. Needless to say, staying curious and being in line with the future of AI opens new frontiers of creative freedom to create something new and unique. As you can see, I try hard to avoid making uncanny morphing effects with AI. Instead, I would like to use this technology to combine all the new tools with my familiar 3D and 2D pipeline. Let’s see what else I can make with all of that. #ai #aiart #digitalart #maxon #maxonc4d #cinema4d #redshift #digitalart #3d #satisfying #visuals #visualisation #future #digitalworld #render #art #editing #contentcreator #aftereffects #motion #motiondesign #animation #krea

One area that I think machine learning is interesting for is giving new life to old assets. This is part of an animation I made when I was first starting out in motion design 5 years ago. Its pretty minimal as I was really just learning the basics at the time, but today I wanted to see what AI could do with it to jazz it up. While the results are frenetic and chaotic, I think it adds something interesting to the piece that was otherwise lacking previously. I like the texture and movement, like watching a living, breathing painting. Playing with this made me wonder - how many artists are using AI to repurpose their old assets? #animation #motiondesign #aiart #comfyui #aftereffects

AI characters - for 3D assets ✅ Connect with me 🤗 👨🚀 ♻️ Feel free to comment, share and repost You can create 3D characters with AI directly from a reference image, but there are few options to check and one is www.viggle.ai Basically you just create a character with AI and animate it. Or course you can create your own video and use it as reference and just swap it with your AI one! Ready? Join the 3D Revolution 🤗👍👨🚀🚀 Remember and Honor the Memorial Day! Image done wiith Midjourney v6 Animation done with Viggle Check my post with Characters: https://lnkd.in/eCd8RH5H …and 3D with CSM https://lnkd.in/esFHZ6Pt

I am so impressed with how AI is progressed in terms of generating images or even Videos. One worry that we have all had is whether in the next two years our jobs as creatives will be overtaken by AI. I have been looking for the perfect AI tool to help me as a motion graphic designer. I came across Prism by Fable. A generative AI feature. Some of the key features in Fable is different animation effects to stylize your animation. I animated the video on the left with adobe after effects then stylized it to the Claymation effect of Prism and voila... I have a whole new animation style. I can't tell you how excited I am to learn and explore Prism and will definitely be including it in my animation journey. #motiongraphics #AI #generativeai #prismbyfable

I do not know how to draw, and I am not an illustrator. But with AI tools, I created this animation without contracting any artists, or paying for any stock vectors. Using a combination of Dall-E, Adobe Photoshop Generative Fill, and Adobe Character Animator, I was able to create this animation of Bill Walton for Stellar Solar. Completely from scratch and without drawing anything. AI is not going to replace you - but it can make you much better at what you do and expand your creative capabilities. Even if you're not artistically talented, like myself, you can use it to make things that you'd have to pay someone to do before, all by yourself, and basically for free. #ai #dalle #adobephotoshop #aiart #billwalton

Using AI for animation -- I motion captured myself doing a "let me take a selfie" moment for a crowd sequence. Still needs some manual editing but not a bad start :-D what do you think? DeepMotion AI gave me the best result out of the box and also had a great editing tool on the web, so I was able to do some tweaks right away before importing to Blender. Tried Plask AI but it did not give me as good a result and was not as intuitive. There are one or two others I've heard of, and I am definitely watching this space as dang, mocap + character animation in general is quite a rabbit hole and very time consuming, which makes it a great use case for innovation with AI. My dream is to be able to mocap a full performing arts piece including facial animations (which by the way, DeepMotion did pretty well in this test). I've seen the Unreal live capture stuff with MetaHumans and that's also something I want to play with -- especially for a realtime performance! Imagine an elaborate character and scene that would be cost-prohibitive to produce IRL, and the performer being able to respond to an audience in a convincing way. That is where I see the benefit of this tech, and I love being able to merge art and tech to create beautiful things. Stay tuned ;-) #artificialintelligence #assistedintelligence #artfulintelligence #art #performanceart #performingart #mocap

🔍 Exploring "AniPortrait: Audio-Driven Synthesis of Photorealistic Portrait Animation" 🎨 Just stumbled upon an interesting read on #PaperwithCode about the AniPortrait framework for creating photorealistic portrait animations driven by audio cues. 🖼️🔊 Here's a quick breakdown: 👉 AniPortrait Framework: A system that crafts lifelike animated portraits using a reference image. It's all about transforming audio into 3D facial representations and then converting them into 2D landmarks for animation. 🛠️ Technological Advancements: Leveraging transformer-based models and a robust diffusion model with a motion module, this approach captures intricate facial expressions and head movements, resulting in natural and diverse animated portraits. 💡 Applications and Flexibility: From facial motion editing to face reenactment, AniPortrait offers significant control over the animation process. Plus, its adaptability shines through in its ability to modify 3D facial representations for various applications. 🔬 Research and Development: The framework showcases superior capabilities in generating animations with impressive visual quality and realism. Excitingly, future work aims to enhance direct audio-to-video generation for even more stunning results. 📚 For those interested in delving deeper or accessing the code and model weights, the authors have kindly provided resources on their GitHub repository. A noteworthy exploration at the intersection of audio and visual synthesis! 🌟#AI #Research #Technology #MachineLearning #MachineLearningEngineer