How Our System works

We're excited to offer you cutting-edge style and AI model implementations that make your projects stand out. To ensure you get the most out of our services, we'd like to explain a bit about how our system works, especially for those using our platform for the first time. This guide is designed to be accessible for both technical and non-technical users.

Understanding the Initial Processing Time

When using our AI platform, you may notice that the first time you generate an image, it takes longer than subsequent generations. This is not a glitch, but a feature of our serverless GPU system:

• Initialization: When you generate your first image, our backend system will prepare a Docker container running on a GPU to handle your request. During this time, you can select the desired GPU option (A10G, A100, or H100) to suit your requirements.

• GPU Options: The A100 and H100 GPUs offer faster processing, but they come with an additional cost. For a more cost-effective solution, you can choose the A10G GPU.

• First Run: The initial image generation will take longer due to the container setup process, similar to organizing a workspace before use.

• Subsequent Runs: After the initial setup, the container is fully prepared, and any further images you generate will be processed much faster.

• Changing GPU: If you decide to change the GPU option in between tasks, the container will need to be restarted. This will result in an extra wait time only for the first generation after the change. Subsequent generations will be faster, as the container is already running.

• Idle Time: To optimize resource usage, the container remains active for 3 minutes after completing a task. If no new tasks are requested during this time, the system will shut down the container to conserve resources.

• Restarting: If you send a new request after the container has been deactivated, the system will automatically restart the process. While this will again take a bit longer for the first image, it ensures efficient resource utilization without compromising on speed for active tasks.

Our serverless GPU system is designed to balance performance and efficiency. By understanding this process, you can better plan your projects and anticipate the time required for image generation, especially when using the system after a period of inactivity or when changing GPU options.

We hope this explanation helps you understand the innovative technology that powers your projects on our platform.