MLMimic - Early Access

Before you buy:

• Please note that MLMimic is currently only available for Windows 10 and 11. Linux support is planned but not currently available.
• An active internet connection is required for license validation. Offline licensing is not available for MLMimic yet.
• MLMimic supports Purchasing Power Parity. Licenses purchased with a PPP discount are only valid for use in the region in which they were originally purchased.
• Only NVIDIA GPUs are currently supported. RTX20 series GPUs are the absolute minimum recommended GPU model.
• MLMimic is still in active development; as such, there may be a few rough edges here and there that will be addressed in subsequent updates.

What is MLMimic?

MLMimic is an alternative, standalone machine learning tool for use in VFX. Allowing artists to take advantage of the power of neural networks in their workflows, regardless of where they choose to do their work.

Why Early Access?

As it stands now, the core features and functionality are in place and solid, with more on the way. 
Early access allows early adopters to benefit from its core functionality now at a lower price point, while helping to guide development priorities with their feedback and receiving free updates and an upgrade to v1.0 when it launches.

What do I get now?

v0.9.1 Features:

• User-friendly, Node-Based UI.
• A variety of Input and utility nodes to facilitate dataset creation, model training and inference.
• Single & Multi-GPU training support.
• Small, Medium, Large and Extra Large Mimic models.
• Training crop sizes from 128x128 pixels up to 1280x1280 pixels.
• Training Hyperparameter presets tuned for various tasks.
• Built-in 'on-the-fly' Data Augmentation options to help improve model learning and generalisation.
• 4 Channel EXR support.
• Sample projects demonstrating MLMimic used for Style Transfer and Subtle Cosmetic tasks.

What do I have to look forward to?

There are plenty of exciting features in development that include:

• New sample projects and video tutorials.
• Scheduled Resolution and Scheduled Crop Size training strategies to help get better results, faster.
• Utility nodes to facilitate the creation of A/B frames inside of MLMimic.
• Pre-trained Mimic models to help speed up A/B frame creation for various tasks.
• New Mimic model varieties.

What do I need to run MLMimic?

MLMimic currently only supports CUDA-enabled GPUs and uses CUDA Version 12.6. It requires a minimum NVIDIA driver version of 560.94. MLMimic should support consumer GPUs from the RTX20 series up to the RTX40 series, as well as Ada-generation RTX and RTX A-series commercial GPUs.