James Jensen

Been experimenting with mixing AI tools and real-time engines to push creative workflows further. This one began with a 3D mesh generated in ComfyUI using Hunyuan3D, then moved into Unreal 5 with custom Niagara scripts. Sound behavior was driven by TouchDesigner, which offered more granular control over the audio analysis for this specific use. There’s something powerful in making different platforms speak the same language. - Background music is In Your Sleep by Kuedo - #AI #3D #genAI #hunyuan #comfyui #unreal #unreal5 #vfx #touchdesigner #CreativeTechnology #RealtimeGraphics #AIGeneratedArt #UnrealEngine #TouchDesigner #NiagaraVFX #ComfyUI #ProceduralArt #GenerativeDesign

15

Top 10 tips ((Cheat Sheet)) for USD Rendering useful for Tech Artists & Pipeline TDs Use it anywhere you need headless, automated, Hydra-based USD rendering without opening Houdini. 1. Basic Render Render a USD file using Karma (default delegate): husk scene.usd 2. Render a Frame Range husk scene.usd -f 1 100 And to point to output path: husk scene.usd -f 1 100 -o /renders/shot1/beauty.$F4.exr 3. Switch Render Delegate Render using Karma XPU: husk -R KarmaXPU scene.usd Render with Storm (fast preview): husk -R HdStorm scene.usd And you can add any Hydra delegate (Arnold, RenderMan, 3Delight if installed). 4. Override Render Settings Change a render var (AOV) directly: husk -set camera /path/to/cameraPrim Override resolution: husk scene.usd -xres 1920 -yres 1080 5. Render Specific Camera husk scene.usd -C /World/camera1 6. Add AOVs Render with ego-defined AOV (example: nrm, depth): husk scene.usd -A nrm -A depth -o /renders/aov.$AOV.$F4.exr 7. Generate a Preview Movie husk scene.usd -f 1 120 -o preview.$F4.jpg Use ffmpeg afterwards: ffmpeg -r 24 -i preview.%04d.jpg preview.mov 8. Progressive Rendering Good for quick review: husk scene.usd -progress 9. Use Husk in PDG Inside a ROP Fetch TOP or USD Render TOP: Set “Command” to: husk @usdfile@ -f @frame@ -o @output@ PDG distributes to farm automatically. 10. Debug Mode husk scene.usd --verbose or husk scene.usd --trace #Houdini3D #USD #Husk #KarmaXPU #PipelineTD #TechArt #Proceduralism #PDG #Hydra #VFXPipeline #RenderFarm #CGWorkflow #OpenToWork

23

6 Comments

Using AI with a light touch can make some cool stuff! 1. Get a depth estimate of a historic photo from AI. 2. Turn it into a "pie plate" 3D model. 3. Spend as much time as the project allows to clean up some mess, add some FX and compositing touches. 4. Profit? Fun Way to advance "The Kid Stays in the Picture Effect" and add subtle life to our crucial history. #AI #documentary #motiongraphics #vfx

7

1 Comment

Aaaand AI again. I’ve been playing around with the recent Sparc3D paper, and… yeah. This might be a real turning point in 3D reconstruction. Imagine taking a single 2D image and getting back a high-res 3D mesh - without needing LiDAR, photogrammetry rigs, or even multiple angles. Sounds sci-fi, but this new method actually delivers surprisingly clean geometry and rich surface detail. Pipeline (ultra-simplified): ▪️ Input - one or more RGB images (even just one is enough) ▪️ Processing - Sparc3D model processes the image and infers dense geometry (volumetric + surface) ▪️ Output - high-res mesh ▪️ Export to Blender / ZBrush / Unity / Unreal - done. Photogrammetry vs Sparc3D Photogrammetry (e.g. Metashape or Reality Capture): ▪️30-200 photos from different angles and accurate camera setup ▪️Lighting control: needed ▪️Geometry fidelity: high ▪️Texture reconstruction: yes ▪️Hardware requirements: Medium–High GPU + RAM ▪️Speed: slow (from minutes to hours) ▪️Ease of use: steep learning curve Sparc3D: ▪️1-3 images ▪️Lighting control: not really ▪️Geometry fidelity: surprisingly good ▪️Texture reconstruction: not yet ▪️Hardware requirements: Mid-tier GPU, ideally ▪️Speed: near real-time ▪️Ease of use: simple input/output (once wrapped) Yes, you still need cleaning-up (topology is not "game ready"). And you still won't get textures. And yes - copyrights are still an open question. But on the other hand - speed, already good surface quality even in places that are hidden in the image, minimum requirements. So, Sparc3D doesn’t replace photogrammetry (yet). But for quick asset prototyping, AR/VR previews, or even indie game dev… this might be exactly what we needed. And again: it's just another tool ;) Paper & Demo Links: 📄 Paper: https://lnkd.in/eamuM2Ch 💻 Demo: https://lnkd.in/em6kZCKu

28

1 Comment

The sun's hidden power: Geomagnetic Storms. The Weather Channel's immersive design team shows how these events test the fragile tech we rely on. #SpaceWeather #TheWeatherChannel #unrealengine

124

15 Comments

Rodin Gen-2 API, image-to-3D tool, is now live in ComfyUI! 🚀 Convert any image into a 3D model in one click—supporting 4× mesh quality, T/A pose, PBR textures, and multiview. 🔥

35

2 Comments

Freepik is pushing AI video quality closer to real production. They have integrated Magnific Precision into their Video Upscaler, focusing on detail, control, and final image quality. This means you can: → Upscale videos up to 4K while preserving real textures → Maintain believable visuals instead of over-processed sharpness The key improvement is where other tools struggle: → Skin texture stays natural → Hair, fur, and vegetation remain detailed → Landscapes keep depth without looking artificial There is also a smart workflow layer built in: → Preview 12 frames before committing to a full render → Validate quality early without wasting time or credits Control is another major upgrade: → Adjust sharpness, grain, and strength → Fine-tune outputs based on style, from clean to cinematic And it doesn't stop at resolution: → FPS can be increased for smoother motion → The result feels better, not just looks sharper This points to a major transformation in AI video workflows. Post-production is becoming more precise, more controllable, and more reliable inside a single system. For startups, brands, or small creative teams, this means: • Higher quality outputs without reshooting • Faster iteration with predictable results • More control over final delivery without heavy post-production stacks → Teams can push existing footage further instead of starting from scratch → Production pipelines become lighter, faster, and easier to scale The way we create and refine video is changing. These are the kinds of tools that will define the next era of production. P.S. Follow me for more AI production insights → https://lnkd.in/e7Z_Fm6q

42

Canva, Affinity, Cavalry. Will you all help a dude out and officially port your software to Linux? Creative EL/Fedora user here who loves making creative work with Blender, Blackmagic Design's Resolve/Fusion, Krita, TVPaint Développement, and SideFX Software's Houdini. I feel like you joining the team would really help turn Linux into a viable full-time option for independent creative professionals looking for a scalable open-source OS environment. Boldly go where Adobe dare not tread! Any other nerds out there feel the same way 🤓 ???

4

How to fix lighting in any shot without shooting a rerun on your production set. In AI production, lighting control can turn to be an immense control problem ⚠️ Because it is more than a creative preference. If the light is wrong, everything downstream gets expensive: → Color grading gets pushed → Skin tones break → Products look flat → ‘Cinematic’ becomes a filter Higgsfield AI's latest release, relight, has made an important impact in production consistency 🎬 Relight uses AI plus depth mapping to behave more like a 3D lighting setup. This means you can control: → Direction → Intensity → Temperature → Color (even via hex) In this case, lighting becomes editable after the shot exists ♻️ Here’s the 3-step guide for integrating Higgsfield’s upgrade in AI production: 1️⃣ Diagnose the lighting failure before you touch sliders → Ask one question: what is missing? Shape (no separation) Detail (no texture) Mood (wrong temperature) → This avoids randomized adjustments and keeps intent clear. 2️⃣ Rebuild direction first, then refine quality: Start with a directional preset (Top, Front, Left, Right, Back, Bottom). Then use the directional pad for natural angles. → Direction creates “cinema.” Brightness only exposes mistakes. 3️⃣ Decide Soft vs Hard based on what you are selling: → Soft light for premium skin, flattering wrap, controlled highlights. → Hard light for texture, contrast, product definition, dramatic mood. → Then fine-tune brightness and color temperature to lock the atmosphere. 🎯 Here is where this shines in practice: Backlit portraits → recover the face without blowing the background. Product shots → add top or side light to create depth and texture. Creative looks → consistent neon, noir, or cinematic palettes across a set. This is exactly the kind of controlled output clients pay for: An output they can visualize and trust. This gets rid of random creative direction and enables repeatable control ☑️

8

🚀 Genesis Verse – Major Development Milestone Achieved MVP Delivered • January 16, 2026 Today marks one of the most significant leaps forward for Genesis Verse → Abyssum Gateway VIP. Multiple core systems came online, the gameplay loop is now fully functional, and the project has officially reached production‑ready MVP status. 🏆 What’s Now Fully Operational 🔹 Web3 Wallet Integration A streamlined, stable MetaMask connection flow is now live. Players can authenticate seamlessly and prepare for future on‑chain features. 🔹 Gateway Universe Portal A new hub page brings external content directly into the experience, including embedded gameplay and community‑facing media. This transforms the Gateway into a true cross‑world access point. 🔹 Inventory System Players can now view and manage four core resources through a clean, responsive UI. Each resource features rarity indicators and dynamic quantities. 🔹 Turn‑Based Combat System A complete battle loop is now active: Three enemy types HP and energy systems Critical hits and ability patterns Victory/defeat flows Combat log for clarity This is the first fully playable gameplay module inside Diesel City. 🔹 ABEX Banking System A collectible banker system is now live, featuring: Ward‑based filtering Rarity tiers Ownership indicators Mint pricing and supply tracking This establishes the foundation for the game’s economic layer. 📹 Video Evidence Captured Two major recordings were completed: A full combat demo A walkthrough of the ABEX Banking interface These clips will support upcoming LinkedIn posts, Itch.io updates, and investor materials. 📈 Performance & UX Smooth navigation across all pages Fast load times Responsive design across devices Clean, glassmorphism‑based UI throughout The experience now feels cohesive, polished, and ready for public eyes. 🎯 MVP Success Criteria Achieved Players can now: Explore Diesel City Manage resources Engage in combat Interact with the banking system Connect a Web3 wallet Access external content portals This completes the core loop for the MVP. 🔜 What’s Next Mobile responsiveness audit Additional gameplay polish Full gameplay loop recording for public release Performance tuning and metadata updates 🟢 Final Status Genesis Verse – Abyssum Gateway VIP is now a production‑ready MVP. The foundation is strong, the systems are stable, and the experience is ready for community testing and professional showcase. #Tags #GenesisVerse #AbyssumGateway #GameDev #IndieDev #Web3Gaming #CombatSystem #InventorySystem #NFTGaming #MetaMask #BabylonJS #UIUX #DigitalWorldbuilding #MVP #DevLog #LinkedInDevLog #SciFiGaming #TurnBasedCombat #GamingEcosystem #ItchioDev #ProductionReady

2

Refining 3D scans in Substance Painter using AI creates compelling results. With every scan I render to 8k or 4k, I often wonder if the pixels are really being utilized. Is it possible I'm just rendering blur to a larger texture? In a gamedev pipeline, that's just wasteful. What's the "perceptual" resolution? Here in Painter, I have a before/after of a basecolor texture from a carved wooden bucket. When you run your base texture through the right workflow, it forces the diffusion process to use the input texture as the base, then refine on top. New fine details are applied and the spirit of the original remains. Old scans can become new again. Broken scans from your outsourcing vendor can be resurrected. Since this is a 2D process, seams will become visible on the mesh across UV shells. Interestingly, the diffusion process made these seams look natural. Look closely at the top rim of the bucket. Open the image in a new tab for a better look.

49

5 Comments

You can go in any direction you want I made this with Luma AI's dream machine to test out Ray 3's animation capabilities. I loved the results! These were make by placing the initial image into the first frame dialog and adding a relevant prompt. Added some music to give it that extra old school feel. Have you tried Ray 3 yet? -------------------------------------- Join hundreds of creatives who follow our ZenRobot newsletter, for weekly GenAI tools and practical tips ♻️ and if you found this post useful, please repost ♻️

112

37 Comments

Smart Config Pro is now live on Fab for Unreal Engine. I built this tool to eliminate repetitive setup work when creating interactive tours, especially in large or fast-moving scenes. With Smart Config Pro, you can convert selected meshes into smart, interactive actors in a single step, manage interaction and widget design through one global settings actor, and keep the entire workflow non-destructive. It’s designed for teams working on configurators, arch-viz, virtual production, or any project where consistency and speed really matter. If you’re spending too much time wiring the same logic again and again, this might be useful. Get it on Fab - https://lnkd.in/d2gVet4w

13

Proud to see how raw facial performance capture transforms into expressive digital characters APPLE ARTS STUDIOS Motion Capture Services. With 15+ years of experience, we continue delivering high-quality motion capture solutions for films, games, and virtual production. #MotionCapture #FacialCapture #VFX #Animation #VirtualProduction

7

How easy is it to create an Ozone Story Tech Subdiv asset in Epic Games Unreal Engine? Let me count the ways :) Right Click on Skeletal Mesh, choose "Cook Ozone Subdivs" Create Ozone Subdiv Info Asset Click "Cook Subdiv" Thats it... and just like that Owl from the 80's trying to get to the middle of the Tootsie Roll Industries Pop you are 3 steps from enjoying smoothy, high resolution character goodness. Rob Di Figlia Duncan Burbidge 😀Tony Bowren Chase Cooper Vaishak Purushothaman Quentin Defrance Atomic Cartoons Tharyn Valavanis Brad Clark

42

2 Comments

When artists integrate AI into their workflow, they can either use Additive or Subtractive AI workflows. With Subtractive AI workflows you create a large swathe of different content. And curate it down to the few items that most closely represent what you may have been going for. With Additive AI, you start from some piece of media, be it a storyboard, concept art, photography or something else that already has a large amount of intentionality behind it. And iterate upwards from there. In my view, both are valid. But Additive workflows are the best for total creative control, resulting in media that is precisely what the artist intended. This is an example of an Additive AI workflow. I start with a clear idea of my character. He's a half Maori head of a security firm. Battle hardened, but from a sophisticated social class. I start by drawing and describing the character and his backstory. I then enhance this drawing in the storyboard mode to speed up the rest of this workflow (optional) And then in Cinematography mode, I bring the character up to a final character sheet. Along the way there were many iterations - over 30 paint overs and modifications, but I'm happy with the result and ready to start filming. Filming itself is done using an AR camera in real or virtual environments, and directed using natural language. Eg. "Mate, can you give it a little more intensity? There's more at stake". You can also use a real actor to drive these performances. This was done in LiveMind Aperture. I'll share more workflows soon.

8

High-end real-time visuals aren’t about pushing complexity. They’re about controlling it. Every technical choice impacts performance, timelines, and long-term flexibility. That’s why I favor texture projection workflows over unnecessary geometric complexity and UV mapping. This preserves visual quality while keeping scenes efficient, scalable, and adaptable, even late in production. Optimization, at this level, isn’t a technical constraint. It’s a creative advantage. Choosing the simplest path is often what enables the most refined result. #RealTimeVisuals #DigitalEnvironments #ProductionWorkflow #UnrealEngine

7