Creative Tools for Animators using Spell
Cadmium is an animation company using machine learning to provide creative tools for artists and animators. Many animation tasks are traditionally done by hand, which can be incredibly time consuming. Cadmium seeks to offload more tedious tasks through automation, freeing up time for animators to focus on the creative process.
Cadmium uses machine learning tools to speed up the animation process in ways that allow animators to scale up their work or create visual effects that haven’t been seen before. Evan Casey, an engineer at Cadmium, says they believe AI is going to have a big impact in how animation is made. If there’s less time being spent on tedious cleanup, that reduces the cost of creating animation in general - for example, enabling a small studio to achieve more than they would otherwise be able to.
The machine learning engineers at Cadmium have found Spell very helpful through having one script that will automatically run everything on the cloud, synced with Git. This is the core part of the Spell product that has been really useful for them. They also are looking forward to using the Weights & Biases integration, which they see as a really good tool for tracking experiments, and more flexible than TensorBoard.
Spell provides the flexibility to use AWS or GCP, enabling Cadmium to make the most of their free credits from AWS Activate and Google Cloud for startups by supporting seamless switches between cloud providers with no interruption to workflow . Team collaboration was made simple and efficient; with the team able to launch experiments and visualize their results quickly. Cadmium appreciates that Spell is web based so that they can share links to their results with ease. Furthermore, they cite good technical support from Spell as being akin to having their own DevOps team — immensely valuable since the DevOps aspects of machine learning can be a huge time commitment.
Evan notes that continuous integration in ML is a really big deal, because bugs are so hard to test. They actually use Spell currently as a form of manual testing - they can test if a code change distorts their experiment results by running a very short experiment on Spell and seeing if it changes the results from the baseline.
In the future, they’re hoping to be able to do this as part of an automatic continuous integration system.
Schedule an in-depth demonstration with a Spell representative to learn how Spell can help streamline and accelerate your machine learning development.