Building Models 10X Faster: Why Scenario Chose Koyeb

Scenario empowers creators to master AI generation, with unmatched control and consistency over asset creation. From training custom models and generating on-brand assets, to editing and upscaling your creations, you can go from ideation to production, all within Scenario. Each tool has been expertly designed to streamline your creative process, build smart workflows, and create faster within their web platform or via their comprehensive API.

The team runs text-to-image, image-to-image, controlled, and custom pipelines across hundreds of thousands of models. These include FLUX1. dev, FLUX1.1 [pro ultra], SDXL, and more, deployed on H100, A100, and L40S GPUs.

"We needed high-performance GPUs with at least 48GB of VRAM. We needed to test and optimize models on different hardware. Storage had to be fast. And we didn’t want to pay for idle compute," says Hervé Nivon, Scenario's CTO.

With Koyeb, we can seamlessly switch between high-performance GPUs, leverage fast NVMe storage, and deploy containerized environments with ease. This allows us to focus entirely on building and optimizing our models instead of managing infrastructure.

Challenge: Testing and Optimizing Models Was Slow

"Our team needed a way to test and optimize our models quickly across different GPUs, without wasting time on infrastructure."

Scenario needed ready-to-use development environments running on high-performance GPUs to develop and optimize their AI models efficiently. The Scenario team started looking for GPU solutions with these requirements in mind:

• High-performance hardware with at least 48GB of GPU VRAM.
• Ready-to-use development environments for instant access to different GPUs with a ready-to-use development environment.
• Fast storage solutions to handle large datasets and model weights without bottlenecks.
• Serverless GPUs to avoid paying for idle resources and experiment across hardware quickly.

Scenario explored specialized providers and hyperscalers but ran into limitations. Hyperscalers were too costly, while other specialized providers lacked critical features like container support or had slow disk speeds. None provided the performance, flexibility, and ease of use they needed.

That’s when Scenario found Koyeb.

Faster Iteration with a Ready-to-Use Dev Environment

"With Koyeb, we can launch the different GPUs we need in seconds. Instant SSH access, CUDA is pre-installed, and everything just works. No setup, no delays. We can dive straight into building and optimizing our models.

Koyeb eliminates infrastructure friction. Scenario's team no longer has to worry about drivers, environment configs, or system updates.

"We bring our own containers, perform our benchmark seamlessly, and iterate rapidly. Everything runs exactly as expected."

With ready-to-use development environments, Herve's team can iterate, test new ideas, and optimize models quickly. The result? More time spent refining models and less time managing hardware.

"Faster iteration means better models. Better models mean better visuals for our users and their players." Team velocity is sped up, cutting down iteration cycles and delivering AI-powered visuals faster.

"We don’t think about infrastructure anymore. We just build."

Testing Across GPUs Without Friction

"Switching GPUs should be seamless, and the development environment should just work. We can switch from an H100 to an A100 in one click."

Koyeb allows Scenario to test models on different GPUs instantly: H100, A100, L40S, and more. Scenario's team can conduct performance benchmarks, optimize their models, and iterate faster—all without infrastructure constraints bogging them down.

High-Performance Storage for AI Workloads

"Fast GPUs mean nothing if storage is slow. Storage should never be a bottleneck."

Scenario’s AI models process massive datasets, requiring fast, reliable storage to keep up. Scenario’s workloads rely on NVMe local volumes for storing large datasets, model weights, and training outputs.

"With Koyeb, disk speeds aren’t a bottleneck. Large datasets load instantly, keeping training and testing running efficiently."

Thanks to Koyeb's high-performance NVMe storage, Scenario experiences faster read/write speeds, resulting in more efficient training and a seamless process from development to production.

Serverless Efficiency Without Idle Costs

"We don’t want to pay for compute we’re not using." Koyeb’s serverless GPUs allow Scenario to scale on demand and only pay for active workloads.

"With Koyeb’s pay-per-use pricing, we only pay for resources when we need them. This flexibility eliminates wasted costs and ensures we’re not tied to long-term contracts or unused capacity."

Before adopting Koyeb, Scenario faced high GPU costs with traditional cloud providers. Koyeb's pricing model offers the high-performance GPUs we need without the financial burden or vendor lock-in.

Seamless Collaboration, Faster Iteration

"Every engineer gets a fully isolated production-like environment in seconds."

With Koyeb, the Scenario team avoids shared resource conflicts and complex infrastructure setups. Each engineer spins up a reproducible environment instantly, enabling faster testing, optimization, and model support.

"We experiment without friction. What works in development works in production."

This streamlined workflow accelerates model optimization and shortens the path from idea to deployment.

Koyeb Features Highlighted

• 🦾 Switching Best-in-Class GPUs in One-Click: Scenario seamlessly runs on H100, A100, L40S, and multi H100 and A100 GPUs powered by Koyeb
• ⚡ Serverless Infrastructure: Automatically scale resources based on usage and only pay for what you need without worrying about idle time or infrastructure management
• 💾 High-performance NVMe volumes: Enables fast read/write speeds for large datasets and model weights and accelerates model training times
• 💡 Bring Your Own Containers: Effortlessly orchestrate and manage containerized environments with Koyeb's powerful container support
• 📦 Pre-configured Instances: No setup required, get SSH access to fully-configured instances
• 👋 Dedicated Support via Slack: Koyeb's team is just one Slack message away if Scenario ever needs

⏱️ 5 Minutes to Deploy Worldwide on Koyeb

Want to deploy AI applications, APIs, full-stack apps, databases, and background workers worldwide in minutes? Sign up and deploy your first service for free.

Build AI-Generated Production-Ready Visuals with Scenario

Scenario empowers creators to create stunning, production-ready visuals with AI. Whether crafting unique game or marketing assets or fine-tuning a custom model, Scenario gives you the tools to bring your creative vision to life.

Ready to level up your game development? Check out Scenario and start creating today!