Measured bakeoffs

Benchmarks you can audit, not slogans.

On 2026-07-22 we ran the same Remotion compositions and a 20-minute 1080p encode across Apple M2 Max, Apple M1 Air hosts, and Cloudflare Containers standard-1 shapes. These pages publish wall-clock, package power where captured, and rate-card cost estimates with the methodology attached.

3 min Remotion · M2 Max111.5 s~1.9 min wall
Same job · CF Containers live2175 s~36.3 min wall
20 min VT encode · M2 Max95.1 sVideoToolbox H.264
Heavy Remotion package power~33 WSoC avg on M2 Max
Claim boundary: numbers below are from named research runs under research/apple-silicon-idle-render-network/benchmarks/results/. Cloudflare dollar figures use published Containers rates × measured wall time (not an invoice). Package watts are macOS powermetrics SoC estimates, not wall-outlet meters. Idle-subtracted energy was not available on the long suite.

Bakeoff reports

Mac render power draw

Package watt and watt-hour samples during Remotion renders, with electricity math at $0.1883/kWh.

Start with the method: How we benchmark Apple Silicon rendering explains the workload pinning, wall-clock, power, projection, and rate-card labels used in every report.

How we present these tests

Every bakeoff page uses the same structure so engineers can skim fast and auditors can dig:

Read the comparison guide Read the methodology Remotion on RenderMac VideoToolbox encoding
Next step: Buyers → buyer hub / waitlist. Hosts → host hub / waitlist.
Remix pages: workload and FAQ views of the same data - 3-minute Remotion · 20-minute Remotion · M1 Air speed · M2 Max speed · Mac vs CF wall-clock · CF Remotion $ · full library.