How Coframe scaled to 100s of GPUs instantly to handle a viral Product Hunt launch.
LLM Inference
All
faceless.video × Runpod ·
Case study
.png){kind=small}
How Faceless.video bootstrapped to $1M+ ARR
Jacob Seeger rewrote Faceless.video's video generation himself, in Python, and then needed somewhere to run it. The app runs on Bubble for the app layer with an LLM for scripting, a stack built to move fast; video generation itself was originally delegated to Cloudinary's API for the MVP. As usage grew, Jacob's own Python generation logic needed a home of its own. He first looked at AWS Lambda, the obvious serverless option, but Lambda's 15-minute execution limit and tighter CPU-memory constraints made it a tight fit for a workload that wanted more headroom than a typical short function call.
$1M+ ARR
bootstrapped independently
50%+
lower video-generation costs after bringing generation in-house
1 founder
built and ran, with a few contractors
From manual Docker builds to git push
Jacob's early deploys were manual: build a Docker image, push it, wire it up by hand. The generation logic itself didn't change in this move, it was wrapped in a handler and packaged in a container to run on Runpod Serverless. Standardizing on a direct GitHub-to-Serverless deploy flow turned the manual process into a single git push.
I believe there is a 15-minute execution time limit for Lambda, which just wasn't flexible enough for what we needed. Even though most of our videos are generated in far less time, we still wanted peace of mind with the flexibility that Runpod's limits offered us.
-- Jacob Seeger, Founder
Room to run without watching the clock
Moving video generation onto Runpod Serverless removed the execution-time ceiling and Lambda's CPU-memory constraints, so jobs get the headroom they need instead of racing a clock. Workers scale up on demand and down when idle, which meant Jacob could keep a pay-for-what-you-use model, and handle traffic bursts reliably as signups grew. This was important, especially as a bootstrapped scaling business.
Runpod is easily one of the most scalable tools that we use in our infrastructure.
-- Jacob Seeger, Founder
.png)
Results
- $1M+ ARR, bootstrapped independently
- 2.5M+ creators signed up to date
- 50%+ lower video-generation costs after bringing generation in-house, per your own estimate
- Built and run by one founder and six contractors
- Deploys now go through a direct GitHub-to-Serverless flow instead of manual Docker builds
I don't totally remember our cost per video back then, but I remember being incredibly relieved that we switched to Runpod. I'd say we likely saved at least 50% on our video generation costs by switching.
-- Jacob Seeger, Founder
Conclusion
Moving video generation onto Runpod Serverless gave Jacob the headroom to keep building the product instead of managing the infrastructure underneath it. Faceless.video runs as a bootstrapped, seven-figure-ARR business today, without raising outside capital or hiring a dedicated infra team.
