// ROKU
Defiant LA worked closely with the Roku in-house creative team to develop motion theory and expand their 2D design into a robust graphic system. When challenged with the parameters of time, budgets, constant requests, promotion needs, quality, and maximum impact, Defiant LA looked to include AI into our process. We built a workflow and created maximum flexibility for future uses. Not only did we create the strategy behind the motion theory, but we constructed the process, engineered the tools, and developed the philosophy to continue Roku’s Sports journey now, and in the future.
3D models were created from 2D footage through our unique multi-tiered AI driven workflow. Then dynamic movement connected scenes together to create a unique
depth to the package.
VIDEO TO MODEL/MOVE AI- MOCAP
We explored Video to Model technology which allows us to output a full 3d rig and motion data that matches the action within the original sport clip. The benefit of this allowed us to create movement that can be applied into the animation rather than a static model. The workflow then applied our color treatment to the rendered 3d model which has added camera movement and lighting, all with TRIPO as our primary AI tool.
This motion theory allows us to GET IN THE ACTION like never before. Well…you could do ALL of this before AI… with lots of rotoscoping, 3D model building and animation. Now our AI workflow speeds up the process and adds a layer of depth to the flat world. Allowing cameras to get exceptionally close to the action.
IMAGE TO MODEL (AI GENERATION)
INPUT IMAGE
There are very little restrictions with the type of image that can be used, and the image does not need to show the full subject, ideally facing towards camera for best results. Our AI could build missing pieces cropped or covered in the original sports clip.
This model can be applied to any type of subject, (character, objects, vehicles, etc.)
OUTPUT 3D MODEL
Outputs come with a 3D mesh and texture maps.
Explored: Meshi AI, Trellis, Meta Open Source
*Note that the limitations with his is the lack of high resolution outputs however these models will only be used as a base to influence the color treatment so high fidelity models will not be as important.
CINEMA 4D
RENDERED 3D MODEL
+ LIGHTING TREATMENT
+ CAMERA POSITIONING
We’re able to import the 3D model from the output and position a camera
and lighting to create any angle and lighting condition of the subject.
POST TREATMENT AND COLOR APPLICATION
POST PROCESS ILLUSTRATIVE EFFECT
Once we have our render, we can apply our treatment and color
application to achieve an illustrative look.
FINAL RESULT
The final step involves adding in the supporting
design elements and camera movement.
TRIPO GENERATED MODEL
The FBX file can be dropped into C4D and the texture map
will need to be updated within the Roku Style Shader material.
Add a point light, camera, and any environmental shapes such
court lines. The Shader can be adjusted as different models will
react slightly differently.
TRIPO GENERATED MODEL
TRIPO GENERATED MODEL
+ ROKU STYLE SHADER
The UV texture mesh can be cleaned up directly in photoshop as well but due to the complex UV map using something like Substance or C4D will generally be easier.
Background elements should be removed to isolate the subject that will be converted into the 3D model.
This helps in removing any overlapping subjects or similar colors in the background which will result in better outputs.
INPUT
OUTPUT
The sports open was crafted to align seamlessly with the overall package while bringing a high-energy, competitive edge to the presentation.
