Sign up
Log in
Sign up
Log in
Live Webinar: June 25th
Join our Builder's Roundtable to learn all about fine-tuning LLMs
Register now

AI art creation platform NightCafe Studio delivers over 400K daily SDXL image creations for users, powered by OctoAI

Blog Author - Pete Sarabia

Oct 19, 2023

4 minutes
Nightcafe Studio, a pioneer in AI art creation, recently worked with the OctoAI team to launch a new personalized image generation application for its customers. Within a period of two weeks, the new application’s usage scaled 4x from its initial daily usage of 100K, to over 400K daily images, with no drop in latency or image creation success rates for end users. During this period, the team also worked with OctoAI to add new checkpoints to further personalize experiences for users. All these were made available to the platforms 100s of 1000s of users at one third of initially anticipated costs and an image creation success rate above 99.99%, on OctoAI.

Nightcafe Studio Goal: Deliver new experiences to “democratise art creation”

NightCafe Studio was founded with a goal to “allow easy creation of personalized art”. Since its founding in 2019, the platform has rapidly adopted and delivered novel art creation experiences building on technology innovations in the space — including generative AI powered image generation. The 100s of millions of images that have already been created on the platform is validation of the value this brings to art creators today.

Nightcafe Studio continues to look for ways to build on the success of their AI art community engagement to deliver new workflows and new personalized art creation experiences. Earlier this year, Nightcafe Studio started working with OctoAI on one such new project — to enable customers to create more customized Stable Diffusion powered images. The vision for the new experience included adding flexibility to select different fine tuning assets like checkpoints and Low Rank Adaptations (LoRAs), and to create their own fine tuning assets. Alongside, it was critical to control costs as they scaled and to ensure a predictable and engaging community-driven experience.

Building and scaling image generation on OctoAI

In the first phase of rolling out the new service, the Nightcafe Studio team was focused on a pilot to ensure that the OctoAI endpoints delivered the required image quality, latency and reliability. The new pilot service was built to use OctoAI’s new image generation endpoints for image creation and made available on the platform. In the first 24 hours after going live, the new service ran just under 100K image creation runs using OctoAI with no errors or latency issues.

In the second phase, the team worked with OctoAI to add desired pre-defined fine tuning assets (checkpoints) for personalization of the outcome of an image creation job. The service would now allow customers to select the desired customization when running image creation. After initial testing, this additional capability was made available on the platform to its users. The new feature saw rapid adoption by users, with daily image creation increasing over 4x within a few days and no perceivable impact to latency or inference success rates.

Speed is key to the AI art experience we deliver. We’ve been able to increase our image generation speeds by 5x with OctoAI’s low latency inferences, and this has resulted in even more usage and growth for our platform!

Angus Russell, Founder @ NightCafe

Building on the success of the new pilot service the team is now working closely to build additional new experiences and capabilities for Nightcafe Studio users including custom fine tuning for object preservation with their own images, advanced image manipulation features, and prompt engineering usability improvements.

400K daily images and growing, with reliable creation, predictable latencies and one third of anticipated image generation costs

The artist’s creative experience is a critical component of the value delivered by Nightcafe Studio, and the image generation latencies and error rates are metrics that the team tracks closely with every new capability that they launch and scale. OctoAI’s new endpoints delivered error rates (<0.005% failed inference calls) and latencies (end to end p50 under 6 seconds and p95 under 15 seconds) which were well within their existing benchmark based targets. These were possible at a total cost of image generation that was under one third of the initial estimate, building on OctoAI’s efficient compute which is delivered to customers as cost savings.

"Speed is key to the AI art experience we deliver. We’ve been able to increase our image generation speeds by 5x with OctoAI’s low latency inferences, and this has resulted in even more usage and growth for our platform!” Angus Russell Founder, Nightcafe. “And we’ve been super impressed with the delivery velocity of new features. We can barely keep up with the pace of new features and enhancements, and we love what these unlock for our customers."

OctoAI Image Gen Solution private preview: Reach out to learn more

The outcomes discussed here were enabled by engineering innovations being added to OctoAI, to add flexibility and reduce costs as you build on generative AI powered image generation. All these and more, will be available in the upcoming OctoAI Image Gen Solution. Our early access customers have already generated millions of images using the solution. If you would like to test the OctoAI Image Gen Solution and evaluate whether it meets your needs, please submit a request for early access.

Your choice of models on our SaaS or in your environment

Run any model or checkpoint on our efficient, reliable, and customizable API endpoints. Sign up and start building in minutes.

Sign Up Today