Start with the website
The site is designed for Cloudflare’s static asset delivery, with a small Worker for protected generation requests. Cloudflare offers free tiers subject to its usage limits. GPU inference is supplied and billed separately.
Then choose how to run the model
Runpod lists dedicated RTX 5090 Pods at $0.99/hour and RTX Pro 6000 Pods at $2.09/hour on its published pricing page checked September 7, 2026. These are pricing references, not recommendations that those GPUs will meet FastH3’s memory or speed requirements. Availability and prices can change.
A dedicated GPU may bill while idle. Serverless workers can scale to zero, but model startup and persistent storage can add cost. Validate memory and end-to-end throughput before committing to a longer rental.
What the calculator leaves out
The estimate includes only GPU hourly charges. Storage, startup time beyond your entered hours, taxes, platform charges and distribution costs are excluded. It does not estimate frames per second or guarantee a real-time broadcast.
How this studio limits generation
Paid requests require an operator token. A persistent counter limits total and daily submitted jobs, including failed or uncertain submissions. Concurrency is restricted. These controls limit this application’s requests; they do not cap a GPU provider’s entire account bill or stop an independently running Pod.
Runpod pricing · Cloudflare Workers pricing · Current H3 Max API pricing