Conductor® for AI (Beta)
LoRA Training and Inference for Studios
Why Conductor® for AI?
Conductor for AI provides studios with streamlined access to a wide range of leading generative models, enabling artistic teams to experiment and iterate quickly through a simple, intuitive interface.
Studios can securely train models using their own IP - without concern that their data will be reused by model creators, while accessing high-performance compute on demand without managing complex infrastructure.
For more technical teams, API access will be available in mid-2026, enabling deeper integration into studio pipelines. Conductor for AI also features a clear, transparent cost structure with no hidden fees.
Simple Model Creation
Add assets, select your base model, configure a few settings, and determine your performance/cost considerations. The cloud infrastructure, orchestration, and execution are all automatically handled by Conductor.
Manage Training Outputs, Run Inference
Dynamically view Epoch outputs as they are generated, selecting best cases for downloading safetensors or running inference directly from the platform.
Key Features
LoRA Training
Studios can easily create a custom model based on a variety of commercial base models. Through our intuitive UI/UX, a user can add internal project assets, adjust some settings, and generate a custom Low Rank Adaptation (LoRA).
Automation
All cloud compute, storage, and orchestration details are automatically allocated for the submission on demand. No need to negotiate reserved instance contracts; simply select and submit. Instantly access countless high-performance GPUs, from A6000s up to H100s, and only pay for the time you spend training.
Visual Review
An artist-friendly approach shows incremental imagery dynamically as it is generated. No need for unfamiliar loss curves, simply focus on the imagery output, and save the version(s) you need.
Automated Inference
Real-time inference available at your fingertips. From the output page an artist can select the best version and run inference against a new prompt. Artists can also change base models and rerun for different looks or adaptations. Artists can also perform advanced LoRA layering for inference, analyzing multiple LoRAs simultaneously.