Benchmarks
Plenty of things matter for video: quality, reliability, cost, how it holds up on a weak connection. Time to first frame is just one of them, the gap between opening a page and seeing the picture, but it is the one a viewer feels right away, so it is where we started. We measure it on the same source video, from a clean browser, over several randomized samples per provider.
We built a small harness to run these in a standardized, repeatable way, and we publish the raw results so anyone can check them. Here is how Rend and Mux compared on the latest run. We will keep adding providers and regions over time.
If you spot a mistake in the methodology, tell us at hello@rend.so.
Regional results
Time to first frame, median of 5 samples per provider. The US run was on 2 July 2026 in a Daytona sandbox. The Europe run was on 2 July 2026 in a Daytona Europe sandbox. Both use Rend's production playback path.
| Region | Provider | First frame (median) | Played | Stalls | Resolution |
|---|---|---|---|---|---|
| US | Rend | 237 ms | 5 / 5 | 0 | 360p (640 × 360) |
| US | Mux | 691 ms | 5 / 5 | 0 | 720p (1280 × 720) |
| Europe | Rend | 633 ms | 5 / 5 | 0 | 360p (640 × 360) |
| Europe | Mux | 913 ms | 5 / 5 | 0 | 720p (1280 × 720) |
Benchmark test video: 0 seconds. Rend selected 360p (640 × 360) in the US run and 360p (640 × 360) in Europe. Mux selected 720p (1280 × 720) in the US run and 720p (1280 × 720) in Europe. In the US run, Rend reached the first frame about 454 ms before Mux on median. In the Europe run, Rend reached the first frame about 280 ms before Mux on median. Both providers had zero stalls in both regions. Rend reported zero browser network errors in both regions.
How the samples spread
Time to first frame for the fastest sample and median in each region.
| Region | Time to first frame | Rend | Mux |
|---|---|---|---|
| US | Fastest sample | 207 ms | 569 ms |
| US | Median | 237 ms | 691 ms |
| Europe | Fastest sample | 616 ms | 454 ms |
| Europe | Median | 633 ms | 913 ms |
How we ran it
- The benchmark test video is 0 seconds long.
- 5 samples per provider in each region, with provider order randomized each round.
- A fresh browser context for every sample: no cookies, no stored state, and the cache disabled.
- A 30 second watch window per sample, timing the first painted frame and counting any stalls.
- Playwright Chromium 149.0.7827.55, run on a Daytona sandbox with the region set to US for the US results.
- Playwright Chromium 149.0.7827.55, run on a Daytona sandbox with the region set to Europe for the Europe results.
- Rend is measured on the production API fast embed path with Tigris-origin media, not the regional play.rend.so edge hosts.
What this does not prove
- This is one video, one region, one browser profile, and one run. It is not a universal claim about either service.
- We did not purge or warm any CDN. Mux serves from its own network and Rend through its production Tigris-origin API media path, each in whatever cache state it happened to be in.
- Encoders, packaging and player implementations differ between the two providers.
- This is Rend's production playback path, not a forced-resolution run. Rend selected 360p (640 × 360) and Mux selected 720p (1280 × 720) in the benchmark viewport.
- The Europe run used a Daytona sandbox requested in Europe; Daytona reported the sandbox target as eu.
- Source file identity is not independently verified beyond matching duration and observable metadata.
The benchmark we trust most is the one you run. Upload a video to both services, press play from wherever you are, and time the first frame yourself.
The raw results
Every number on this page comes from these files, redacted of any secrets before they are published.
US summary JSONUS raw samplesEurope summary JSONEurope raw samples