ANTON PASKAL 👽

New showreel my dudes. It doesn’t include all the cool footage from last year though. Something still hasn’t been released and under NDA, something is still ingoing from non-commercial, because you know - without budgets we working on it in our free time 🤷‍♂️ Anyway, another two big projects will be released this year, so I can’t wait. In 2024, I also suddenly began my journey as an art director. I will try to realize it even more in 2025.

Here’s the secret behind that cinematic look. This camera prompt from the previous video works insanely well across different locations and scenarios. I’m using the classic combo of Nano Banana and Kling here. I feed the camera prompt into both the image generation and the animation stages. I’ve noticed that in Kling, the camera movements turn out way more cinematic and smooth with it, plus you get this really nice film grain in the shot. If you animate without it, it honestly just looks like a basic, generic video - either completely lacking that cinematic vibe or just not having enough of it even with strong starting frame. #generativeai #aivideo #aicinematic #imagetovideo #klingai

«Generate a video with cool dude. Don’t make mistakes» Made with Nano Banana and Kling.

Monochrome obsession 🌚 Visuals generated in Midjourney + Nano Banana. Animated in Kling. Workflow Freepik. @midjourney @klingai_official @freepik #generativeai #imagetovideo #aivideo #aicinematic

I’m so stunned with how Kling handles flowers. Floristic stuff is one of my favorite subjects, even back in Houdini, and I couldn’t be more happy with this results. Also this time I generated the visuals in Flux 2 Pro, not with Nano Banana. I compared over 30 different variants with the same prompts. And the results in 95% of cases... Let’s just say Flux won on that one 🤷‍♂️ #klingai #fluxai #generativeai #aivideo #imagetovideo

Another day, another round of AI experiments 1️⃣ Seedance 1.5 Pro [Start \ End Frame] I wanted to test how cleanly the model handles transitions and preserves composition. Based on the main image, I generated several new angles, and the result was surprisingly good — almost no glitches, the fabric moves nicely, the motion is smooth, and the overall dynamics turned out solid. 2️⃣ Nano Banana + Kling 2.5 — changing the style The version with the moon looked pretty cool, I liked it, but I wanted to see some more variations, plus I had the idea to somehow combine this composition with lava. So I used Nano Banana to change the style and did the animation in Kling 2.5. Watching the outcome was a pleasure — very consistent and good results. 3️⃣ Macro shot and lava simulation — pure chaos I wanted to make a macro shot with realistic lava: a foot stepping on the ground, the ground cracking, lava emerging, all that stuff. I tested a bunch of models and tried different generation methods using Reference Image and Start Image. The result was disappointing. None of the models understand lava physics whatsoever. The flow, texture, viscosity. Almost everywhere it looks like some type of water, with random splashes. I ended up with about 30 generations, and maybe a third were even remotely usable — the rest went straight to the trash. Feels like this part needs motion control or sum like that to make at least a simple animatic in Cinema/Houdini with a proper foot movement, and then try to restyle it in something like Luma Modify. That would probably give a better chance of success.

I spent some time testing out different workflows both text-to-video and image-to-video to see what these models are actually capable of right now. I specifically tested the most common challenges in commercials like: object morphing, object reveal, extreme close-ups, dolly zoom and cloth simulations. Some results were mind-blowing, some were pretty ass. It’s fascinating to see the strengths and weaknesses of each model when you throw complex tasks at them.

Exploring the creative power of AI engines. In this post, I compare results from two image generators — Midjourney and NanoBanana. Same prompts, different vibes!

AI render test 001 🪷 Today I tested several different AI rendering options on a shot for which I once simulated this flower in Houdini. First variant I fed it our reference styleframe rendered in Octane. So you’re still wasting time on lookdev – basically just cutting out the actual render time. The result’s decent, but not exactly flawless. I also tried generating the visuals from scratch to skip the lookdev phase. Same story: turned out pretty good, but still not perfect. You’ve definitely got to do some manual cleanup either way—masking out random shapes in the background or dealing with artifacts in the first second. You end up having to trim and retime it, which could technically mess up the flow. I used Ray 2 and Kling to process the animatic. Overall, it feels a lot like working with a regular render engine. You’re sitting there tweaking the style/look just like you’d tweak shaders and lighting in a classic workflow. The only difference is the output is way more unpredictable right now, and the generations cost money. Everything you see on screen probably set me back like $15-20 worth of tokens.

Continue to exploring blooming techniques in Houdini 🪷 • #houdini #houdinifx #redshift #sidefx #motiongraphicsdesign

Kaspersky SD-WAN To be honest, it wasn’t an easy challenge to create a concept and a video that clearly and beautifully explains things like centralized network management, traffic routing, security policies, and intelligent tunnels. But challenges like this are exactly what inspire us. We built a car from scratch, designed it, and hit the road. Production: Tribute CG Creative Director: Vasilii Guliaev Art Director: Anton Paskal Producer: Levon Amirkhanian Project manager: Darina Chemeris Account manager: Robert Porsugian Storyboard: Polina Shvez CG artists: Maria Eliseeva CG artists: Sasha Key CG artists: Artur Shafeev CG artists: Infaith CG artists: Alexandr Pinchuk Vehicle design: Denis Plekhanov UI design: Denis Lugansky Animation: Evgeny Elsukov Modeller: Konstantin Zalazaev Colorist: Andrey Khudyakov Sound: Evgeny Kanya rep.by neu team

This is cool, Anton Paskal! 😍 @antonpaskal worked on this knitting effect in Houdini. He is a 3D Motion Graphics Designer, specializing in animation and simulations. His main tools are Houdini and Redshift and he usually does CGI for commercials, and non-commercial art projects. Join our Discord to meet many Houdini artists worldwide. Link in bio. #sidefxhoudini #sidefx #procedural #3d #houdini #cgi #design #animation #motiondesign #fx #3Dart #digitalart #vfx #render #houdinifx #simulation #redshift #knit