Ramon Blake Guthrie

This is a recommended watch, there are a lot of great, insightful chats, no matter what your views are anti, pro, or just curious about AI.

💡 I had an opportunity to explore the capabilities of 𝗜𝗖-𝗟𝗶𝗴𝗵𝘁, in the world of photography and digital imaging, adjusting fixed lighting or simply wanting to try out different environment conditions often demands extensive reshooting, rerendering, or post-production work, consuming valuable time and resources. IC-Light, could be a game changer and a lot more impressive than I first thought, and change how we approach these challenges. ✨ 𝗜 𝗵𝗮𝘃𝗲 𝗯𝗲𝗲𝗻 𝗲𝘅𝗽𝗹𝗼𝗿𝗶𝗻𝗴 𝗽𝗼𝘀𝘀𝗶𝗯𝗹𝗲 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲𝘀: 🏷️ 𝗪𝗵𝗶𝘁𝗲 𝗟𝗮𝗯𝗲𝗹 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝘀: Generate product images with varied lighting and backgrounds, enabling businesses to experiment with different aesthetics seamlessly. 🎨 𝗖𝗼𝗻𝗰𝗲𝗽𝘁 𝗗𝗲𝘀𝗶𝗴𝗻: Rapidly produce concept images, allowing designers to explore diverse ideas without early committing to a final design. ⚡ 𝗥𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗜𝘁𝗲𝗿𝗮𝘁𝗶𝗼𝗻: With the right skillset, you could set up a workflow to quickly iterate on design ideas, making adjustments and seeing results instantly (as seen in my video). 🌟 𝗖𝗿𝗲𝗮𝘁𝗶𝘃𝗲 𝗥𝗲𝗶𝗺𝗮𝗴𝗶𝗻𝗮𝘁𝗶𝗼𝗻: Transform existing photos by altering lighting, backgrounds, and other elements for unique and captivating visuals. 🎥 𝗣𝗿𝗲-𝗩𝗶𝗿𝘁𝘂𝗮𝗹 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻: Ideal for event planners, small studios, and filmmakers to visualize complex scenes, optimizing resources before full-scale production. 🏛️ 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗮𝗹 𝗣𝗿𝗲𝘃𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻: Experiment with lighting and environmental factors, enhancing communication of concepts to clients and stakeholders. In my latest video, I created a workflow in ComfyUI to demonstrate some of the possibilities of IC-Light by adding additional features to better control lighting and showcase different style looks. This powerful tool enables you to unlock new creative possibilities and experimentation. 𝗪𝗵𝗮𝘁 𝗱𝗼 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝘄𝗼𝘂𝗹𝗱 𝗯𝗲 𝘆𝗼𝘂𝗿 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲𝘀? 📄 𝗣𝗹𝗲𝗮𝘀𝗲 𝗡𝗼𝘁𝗲: All my tests shown are fully tested and created by me, built on workflows I have fully developed and optimized to run on my local PC. 🔗 Links: 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝗱 𝗕𝗲𝗵𝗮𝗻𝗰𝗲: https://lnkd.in/ebE7HGHu #ICLight #Relight #GenAI #ComfyUI #VisualReimagination #CreativeInnovation #LookDev #Workflow #CreativeTechnologist

🔮𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁- Lifelike Animations from a single photo 💡The Quickest way of thinking of this technology, is to Imagine the talking portraits from Harry Potter's castle coming to life! ✨ Special shoutout to Douglas McGinness, Leo Kadieff, and Beck Nadir for their great work with 𝗖𝗼𝗺𝗳𝘆𝗨𝗜 be sure to check out their posts. 🧠 𝗡𝗼𝘄 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗧𝗲𝗰𝗵 𝗧𝗮𝗹𝗸: LivePortrait Animation creates lifelike video and realistic motion from a single image. The AI places invisible dots on key facial features like eyes, nose, eyebrows, and mouth, enabling intelligent, efficient, and realistic facial movements. Trained on about 69 million high-quality frames of real, cinematic, dynamic videos that show maximum expressions, this gives the AI model to excel at tracking and matching any motion in a video. It's the fastest motion-tracking AI model for consumer GPUs. 🧪 𝗘𝘅𝗽𝗹𝗼𝗿𝗶𝗻𝗴 𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁: Since Friday, I've been testing LivePortrait's capabilities, running multiple images in Comfy, focusing first on its limitations to understand its strengths. ✨ 𝗪𝗵𝗮𝘁 𝗜𝘁 𝗘𝘅𝗰𝗲𝗹𝘀 𝗔𝘁: • Works with faces turned to the side or slightly tilted. • Tracks faces perfectly, even with glasses. • Tracks faces from various distances, including blurred ones. • Can track multiple faces in one video. • Supports tracking for paintings, statues, cartoons, anime, and 3D faces. 🚫 𝗪𝗵𝗮𝘁 𝗜𝘁 𝗦𝘁𝗿𝘂𝗴𝗴𝗹𝗲𝘀 𝗪𝗶𝘁𝗵: • Only works with humanoid faces; non-human faces cause errors. • Short hair tracks best; long hair can cause distortions. • Struggles with harsh lighting and shadows, leading to perspective distortions. • Retargeting breaks with excessive head movement like head turning or shaking movement. • Can't animate non-humanoid faces or pets. The code is public, and this is just the beginning. Expect more iterations, improved workflows, and new use cases soon. 📄 𝗣𝗹𝗲𝗮𝘀𝗲 𝗡𝗼𝘁𝗲: All my tests shown are fully tested and created by me, built on workflows I have fully developed or optimized to run on my local PC. 🔗 Links: 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝗱 𝗕𝗲𝗵𝗮𝗻𝗰𝗲: https://lnkd.in/eM9WTEbS 𝗖𝗼𝗺𝗳𝘆𝗨𝗜: https://lnkd.in/eR2EUVk7 𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗮𝗽𝗲𝗿: https://lnkd.in/eWf8PzFZ 𝗦𝘁𝗮𝗯𝗹𝗲 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻: Stability AI #LivePortrait #AIVideo #AIFilm #GenAI #ComfyUI #StableDiffusion #VisualReimagination #CreativeInnovation #LookDev #Control #Animation #MotionTracking #TechInnovation #AI #Innovation #FaceTracking #ImagetoVideo #VideotoVideo #AIArt #AINews #AITools

💡 I had an opportunity to explore the capabilities of 𝗜𝗖-𝗟𝗶𝗴𝗵𝘁, in the world of photography and digital imaging, adjusting fixed lighting or simply wanting to try out different environment conditions often demands extensive reshooting, rerendering, or post-production work, consuming valuable time and resources. IC-Light, could be a game changer and a lot more impressive than I first thought, and change how we approach these challenges. ✨ 𝗜 𝗵𝗮𝘃𝗲 𝗯𝗲𝗲𝗻 𝗲𝘅𝗽𝗹𝗼𝗿𝗶𝗻𝗴 𝗽𝗼𝘀𝘀𝗶𝗯𝗹𝗲 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲𝘀: 🏷️ 𝗪𝗵𝗶𝘁𝗲 𝗟𝗮𝗯𝗲𝗹 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝘀: Generate product images with varied lighting and backgrounds, enabling businesses to experiment with different aesthetics seamlessly. 🎨 𝗖𝗼𝗻𝗰𝗲𝗽𝘁 𝗗𝗲𝘀𝗶𝗴𝗻: Rapidly produce concept images, allowing designers to explore diverse ideas without early committing to a final design. ⚡ 𝗥𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗜𝘁𝗲𝗿𝗮𝘁𝗶𝗼𝗻: With the right skillset, you could set up a workflow to quickly iterate on design ideas, making adjustments and seeing results instantly (as seen in my video). 🌟 𝗖𝗿𝗲𝗮𝘁𝗶𝘃𝗲 𝗥𝗲𝗶𝗺𝗮𝗴𝗶𝗻𝗮𝘁𝗶𝗼𝗻: Transform existing photos by altering lighting, backgrounds, and other elements for unique and captivating visuals. 🎥 𝗣𝗿𝗲-𝗩𝗶𝗿𝘁𝘂𝗮𝗹 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻: Ideal for event planners, small studios, and filmmakers to visualize complex scenes, optimizing resources before full-scale production. 🏛️ 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗮𝗹 𝗣𝗿𝗲𝘃𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻: Experiment with lighting and environmental factors, enhancing communication of concepts to clients and stakeholders. In my latest video, I created a workflow in ComfyUI to demonstrate some of the possibilities of IC-Light by adding additional features to better control lighting and showcase different style looks. This powerful tool enables you to unlock new creative possibilities and experimentation. 𝗪𝗵𝗮𝘁 𝗱𝗼 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝘄𝗼𝘂𝗹𝗱 𝗯𝗲 𝘆𝗼𝘂𝗿 𝘂𝘀𝗲 𝗰𝗮𝘀𝗲𝘀? 📄 𝗣𝗹𝗲𝗮𝘀𝗲 𝗡𝗼𝘁𝗲: All my tests shown are fully tested and created by me, built on workflows I have fully developed and optimized to run on my local PC. 🔗 Links: 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝗱 𝗕𝗲𝗵𝗮𝗻𝗰𝗲: https://lnkd.in/ebE7HGHu #ICLight #Relight #GenAI #ComfyUI #VisualReimagination #CreativeInnovation #LookDev #Workflow #CreativeTechnologist

🔮𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁- Lifelike Animations from a single photo 💡The Quickest way of thinking of this technology, is to Imagine the talking portraits from Harry Potter's castle coming to life! ✨ Special shoutout to Douglas McGinness, Leo Kadieff, and Beck Nadir for their great work with 𝗖𝗼𝗺𝗳𝘆𝗨𝗜 be sure to check out their posts. 🧠 𝗡𝗼𝘄 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗧𝗲𝗰𝗵 𝗧𝗮𝗹𝗸: LivePortrait Animation creates lifelike video and realistic motion from a single image. The AI places invisible dots on key facial features like eyes, nose, eyebrows, and mouth, enabling intelligent, efficient, and realistic facial movements. Trained on about 69 million high-quality frames of real, cinematic, dynamic videos that show maximum expressions, this gives the AI model to excel at tracking and matching any motion in a video. It's the fastest motion-tracking AI model for consumer GPUs. 🧪 𝗘𝘅𝗽𝗹𝗼𝗿𝗶𝗻𝗴 𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁: Since Friday, I've been testing LivePortrait's capabilities, running multiple images in Comfy, focusing first on its limitations to understand its strengths. ✨ 𝗪𝗵𝗮𝘁 𝗜𝘁 𝗘𝘅𝗰𝗲𝗹𝘀 𝗔𝘁: • Works with faces turned to the side or slightly tilted. • Tracks faces perfectly, even with glasses. • Tracks faces from various distances, including blurred ones. • Can track multiple faces in one video. • Supports tracking for paintings, statues, cartoons, anime, and 3D faces. 🚫 𝗪𝗵𝗮𝘁 𝗜𝘁 𝗦𝘁𝗿𝘂𝗴𝗴𝗹𝗲𝘀 𝗪𝗶𝘁𝗵: • Only works with humanoid faces; non-human faces cause errors. • Short hair tracks best; long hair can cause distortions. • Struggles with harsh lighting and shadows, leading to perspective distortions. • Retargeting breaks with excessive head movement like head turning or shaking movement. • Can't animate non-humanoid faces or pets. The code is public, and this is just the beginning. Expect more iterations, improved workflows, and new use cases soon. 📄 𝗣𝗹𝗲𝗮𝘀𝗲 𝗡𝗼𝘁𝗲: All my tests shown are fully tested and created by me, built on workflows I have fully developed or optimized to run on my local PC. 🔗 Links: 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝗱 𝗕𝗲𝗵𝗮𝗻𝗰𝗲: https://lnkd.in/eM9WTEbS 𝗖𝗼𝗺𝗳𝘆𝗨𝗜: https://lnkd.in/eR2EUVk7 𝗟𝗶𝘃𝗲𝗣𝗼𝗿𝘁𝗿𝗮𝗶𝘁 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗮𝗽𝗲𝗿: https://lnkd.in/eWf8PzFZ 𝗦𝘁𝗮𝗯𝗹𝗲 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻: Stability AI #LivePortrait #AIVideo #AIFilm #GenAI #ComfyUI #StableDiffusion #VisualReimagination #CreativeInnovation #LookDev #Control #Animation #MotionTracking #TechInnovation #AI #Innovation #FaceTracking #ImagetoVideo #VideotoVideo #AIArt #AINews #AITools

💡 If you are a 3D Artist and ComfyUI user, give Soojung Kwon PBR custom node a try!

Thanks to the Upscale Conf team, for hosting such an insightful London event, now I'm fully aware that a lot of people's timelines have been tainted by bad AI slop videos on X, Instagram, TikTok, LinkedIn, Facebook and Bluesky Social and even though I have been in the AI space for over 3 years now, I was surprised by the talented use AI models to produce new forms of storytelling at a high quality by one person, if I knew in advance how great this was going to be I would brought along more people in the creative and motion image space. I would strongly recommend trying to find a stream for Upscale Day London Event.💡 Special thanks to Ikenna Mokwe ⭐️ , Faye Garland, Neill Torbit, Max Winston, 🌟Rourke Sefton-Minns, Joaquin Cuenca Abela and Freepik

💡 It's definitely a Deepseek-R1 level moment for Image generation. Z-Image-Turbo is hands down the best most accessible and runnable, lightweight image generation model (Systems with only 8GB vram), running at only 6B parameters, which is half the size of Flux.1 Dev 12B parameters model with far more superior quality and prompt understanding. This model can even run on most current Macs. 🧬 Because of this models low requirements even training LoRA models locally is possible, if you're only just starting out in AI image generation or training LoRAs this model is an east choice. Thanks to the teams at Tongyi Lab & Alibaba Group #AI #ZImage