Image, video and 3D
Turn your media generation pipeline into a service that scales to infinity.
Expand limitlessly, reduce to none
"Automatically scale to numerous GPUs as required, and then to zero when idle. Avoid setting up autoscalers or paying for inactive clusters!
GPU Containers
Ultra-quick initial launches
Quickly activate containers with sizeable media models in seconds, enabling agile scaling while minimizing idle GPUs
From single task to workflow
Dibtrun enables linking functions with varied hardware needs and image specifications. Transition from a GPU endpoint to a complete pipeline with CPU-only container-based pre- and post-processing
Streamlining LoRA
Low-rank adaptation (LoRA) allows crafting fine-tuned models as compact adapters for base models.
Dibtrun parametrized functions simplify creating applications for inferring various LoRA adapters. Volumes and immediately have them ready to go for inference.