Weβve never tried a POV shot so we had one of our artists try it out, flying through the skyβ¦ hereβs the result from Cinema 4D and Cinema 4D + StyleFrame
The way the rendering caught all the rock structures and mountains was quite impressive to seeβ¦
Try Restyle Video free: https://t.co/MFvAHB53qQ
Hereβs how the cute characters from our last post came together. They were turned into 3D with Restyle Video, but the base animations were too good not to use, so those are what you saw!
We used 4 keyframes to generate 120 frames of video with AI, while *precisely* controlling when each image appears. Steal our entire process:
(1) Start with 4 still images. One for the beginning of the shot, one for the middle, and one for the end.
(2) Place them on the timeline in StyleFrame. You can position each image at the exact moment you want it to appear in the video.
(3) Adjust the timing. Move the middle image earlier, and the model has fewer frames to transition from the first image to the second. Move it later, and it has more frames to make that transition. That one adjustment changes the pacing of the entire shot.
(4) Once you're happy with the timing, hit render. The model generates the other 116 frames to connect everything into a continuous video.
Four images, 116 generated frames, and a complete 120-frame sequence.
That's what I find so exciting about building creative tools around diffusion models. Instead of describing every movement in a prompt and crossing your fingers that the model gets the timing right, you can make those decisions visually, using controls you're already familiar with.
You decide what happens and when.
The model figures out how to get there.
That's how the magic of video diffusion works.
P.S. If you're an animator or motion designer and want to try this workflow, send me a DM! Happy to share :)
Our talented artist - Ethan - strikes once again with making a short animation based on characters from his past illustrations.
We asked him to create a longer story this time and this is the 30 second result.
Animated in Frames to Video π
Cleaning up shots in post can be a very time consuming process, especially with video.
Our new Edit Video tool changes that. Select specific parts of your input video with our annotation feature, describe what you want changed, and have it happen instantly in minutes. The other parts of your video remain untouched.
Try it free: https://t.co/MFvAHB53qQ
A great software engineer can now run 10 AI coding agents at once. I think we're getting dangerously close to seeing what the equivalent looks like for animators, motion designers, and VFX artists.
Animation is going to have its Cursor moment in the next 2 to 3 years.
Think about what a really talented engineer can accomplish today. They can delegate work to multiple AI agents simultaneously while focusing on the decisions that require deep expertise.
I think we're about to see something similar happen with animation, motion design, and VFX.
Video diffusion models are improving rapidly, and we're approaching a point where one talented animator could:
(1) Explore 10 different visual directions for a scene simultaneously.
(2) Turn rough animation blockouts into finished-looking shots without rebuilding everything in 3D.
(3) Generate variations of lighting, environments, and effects while working on the next scene.
And we're still at the very early. Only a tiny fraction of the animation, motion graphics, and VFX produced today involves video diffusion.
If software engineering is any indication of what's coming, I think the amount of work a single animator or small creative team can produce is about to increase dramatically.
We're going to see creative projects that would've required entire studios become possible for teams of just a handful of people.
Anyone else thinking about this?
Massive win! 15,000+ video editors, animators, motion designers, and VFX artists have signed up for StyleFrame. Along the way though, we've noticed something strange during onboarding.
They'll get excited about four or five features, then point to one they'll never touch. Almost always, it's the feature designed to do the very thing they've spent years mastering.
An animator, for example, doesn't necessarily want AI to handle their animation because a) they've spent years honing that skill, and b) it's probably their favorite part of the process.
But they'll happily use AI for work outside their specialty:
- Coloring an existing animation
- Experimenting with different visual styles
- Adding VFX that would otherwise require a specialist
We've seen this firsthand, and it's changed how we think about building the product. We're paying WAY more attention to the gaps between specialties, where AI can help professionals handle work they don't necessarily have the skills or time to do themselves.
I think that's where AI has the most potential in creative work. Give a great artist everything they need to do their best work, and leave the craft to them.
A few months ago, going from the blockout on the left to the animation on the right would have taken an artist hours or even daysβ¦.
It now can be done in minutes.
Using Restyle Video to speed up 3D PreViz is a huge unlock for visual alignment.
By adding a single reference image, Restyle Video can transform the roughest of blockouts into rendered outputs. All of this, while respecting the direction and timing from your blockouts.
Try it free: https://t.co/MFvAHB53qQ
Removing an object from a video is easy with AI. Removing it without destroying everything behind it will leave VFX artists pulling their hair out trying to fix the mess AI created.
Take the example in this video. Two people are walking their bikes through a shot. We remove the bikes, but that leaves us with another challenge: recreating everything they were blocking, including parts of the people and the background. And because the people are moving, those reconstructed details have to look natural and stay consistent across every frame. Otherwise, you end up with distorted images, flickering, and more manual cleanup for the VFX artist.
That's exactly what we focused on when building the cleanup tool in StyleFrame. With it, you can:
(1) Remove multiple people or objects from the same shot.
(2) Separate foreground and background elements so you can work on them independently.
(3) Reconstruct missing areas between overlapping subjects, rather than just filling in the space around them.
(4) Preserve the video's original aspect ratio, duration, and timing, frame by frame.
Removing an object shouldn't leave artists with hours of extra work. We want VFX artists to spend less time fixing distorted backgrounds and flickering frames, and more time getting the shot right.
In the next 5 to 10 years, every single pixel anyone sees, anywhere, will have been processed by diffusion in some way.
Right now, I'd estimate it's less than 1%, but the change is inevitable.
Much of the math behind modern 3D graphics dates back to the 1970s and 80s. Pioneers like Ed Catmull established foundations that rendering engines still build on today, with vastly more computing power and decades of improvements.
Diffusion introduces a different way to generate images. It starts from noise and gradually resolves it into something recognizable, opening up possibilities beyond traditional rendering techniques.
I think we'll eventually see it everywhere, some examples:
Camera filters:
> Diffusion-powered effects so subtle most people won't even notice them.
Animation and VFX:
> Another way for artists to transform and refine shots.
Real-time graphics:
> Diffusion running on our phones, enabling new AR experiences.
Whatβs important to keep in mind is that a new rendering engine doesn't replace the people who use it. The artists with the most control over diffusion will make the best work with it, just as the best work has always come from the people who knew their tools best.
There's still an enormous amount of potential ahead, and I believe professional artists will be the ones who show us what this technology is truly capable of.
Storyboard to Video Student Film Challenge
One of the hardest things to get right in Video Generation is character consistency across scenes. Not just one great shot, but a character that actually holds together across a whole story.
We put this to the test with a student challenge, and one of the films that came back felt like our childhood days watching cartoons π₯²
The characters held consistently across every scene with a real narrative arc carrying it from start to finish. That same consistency is starting to show up in product work where something that needs to look the same shot to shot now actually does. Much more to come!
A typical shot can involve 3 to 5 different jobs in a traditional VFX pipeline. With StyleFrame, one artist can handle the entire process.
This is a detailed example we created using footage of an actor against a green screen. Instead of our tool treating the entire video as a single pass, we designed the workflow around the way VFX artists have worked for decades:
Separate the shot into layers.
Make the necessary changes to each layer.
Composite everything back together into the final shot.
For context, whenever we develop a new tool, we start by looking at how professionals already accomplish the task. Then we figure out where AI can take on some of the work without taking creative control away from the artist.
For video editors and smaller creative teams that don't always have dedicated VFX resources, we're excited about what this makes possible.
And for experienced VFX artists, we're building tools that fit the workflows they've spent years mastering.
Much more to come!
Our go-to workflow for 3D character animation: Womp + AfterEffects + Glyf
Here's the process:
- Sketch a reference
- Build the first frame in Womp
- Create vector version for animation
- Animate the 2D base in After Effects
- Bring into Glyf for final treatment
Each animation had its own optimal tracking method - the yellow character worked well with Line Art mode, while the smaller dancing characters responded better to the 3D Depth map control.
Always experimenting with different approaches to see what works best for each shot. Shoutout to Hassan Deraoui for his help on the project!
Rendering a 3D flag animation using Blender + Glyf
We created a simple flag simulation in Blender and then applied a style reference using Glyf to render the scene with realistic lighting and materials.
We were mainly curious to see how the lighting in the scene would turn out since the style reference created a frame where the sun was occluded by the flag.
Glyf interpreted the lighting physics correctly with translucency on the flag as the sun rays bled through the fabric texture, and added a small lens flare when the sun was fully visible.
It never ceases to amaze me what is possible when you use video generation as a rendering layer on traditional 3D techniques.
Lighting this scene from scratch would have taken hours if not a couple days. This method at least provides a better starting point for the production process.
Link to workflow breakdown in comments
#3danimation #blender #lighting #glyf
Studios don't need AI to replace animators. They need it to handle tedious production tasks.
William P used Glyf to colorize this 2D robot animation. He wasn't generating new assets, but transforming his black and white animation while maintaining creative control.
Enhanced line art detection handles the technical execution so you can focus on creative decisions. The real value isn't just speed - it's creative freedom. When you can test multiple color treatments quickly, you devote more energy to refining the vision.
Link to workflow in comments!
Video2Video style transfer using Cinema4D + Glyf
A Cinema4D looped animation of procedural geometries morphing and revolving quickly style transferred with high level of consistency.
Styles applied:
- Raw meat
- Red jelly
- Wood
- Chocolate
- Leather
- Metal
- Cheese
- Ice
Glyf allows to assign up to 3 reference frames across the base video timeline, making it easy to maintain and drive the style transfer even when the object is completely changing shapes in a fast and dynamic way, by assigning keyframes as visual control checkpoints for the video generation.
I split the base video into 2 chunks in order to assign 4 total frames across the base video: 0, 40, 98,156 across a 198 frames base video (0, 40, 98 for the first, 98, 156 and 0 for the second, thus closing the loop on the first frame).
Reference frames were created using Stable Diffusion 1.5 i2i in ComfyUI with Canny and Depth ControlNet and IPAdapter for maximum control over style and composition. Ipadapter images for each style are shown in the video.
Once loaded into Glyf I achieved motion consistency using its own Depth ControlNet at 80% intensity.
Link to the workflow in comments :)