video_tool.py render: Is there a way to increase speed

We tried to use video_tool.py to create videos, and found that the recording time is very long and errors are often reported (for example “chunk_worker.bundle.js 54 Error retrieving chunk [object Object]:3,3,0: Error: Expected at least 8 bytes”).

We tried to modify the --refresh-browser-timeout parameter, and the result was obtained.
The final command we run is “ render --hide-axis-lines --width 1920 --height 1080 --refresh-browser-timeout 6000 --jobs 6 --scale-bar-scale 2”.

This data consists of 1 obj file (35 m) and 117 pieces of precomputed annotation ELLIPSOID(total 170 m). Each piece was created as one chunk, with 10k-60k points (a total of 4.2 m points, with an average of 36 k points). There are 5 neuroglancer-url of different states and 2s as transition-duration.

The program ran for 7h on machine 1 and 8h on machine 2。

Machine 1 : Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz (10 cores 20 threads) *2 , 126G Mem, neuroglancer 2.22, python 3.8.8 ,Chrome 93.0.4577.63, ChromeDriver 93.0.4577.63， Nginx-based Web file server, http1.1 by default

Machine 2 : Intel(R) Xeon(R) CPU E5-2650 v4 @ 2.20GHz (12 cores 24 threads) *2 , 252G Mem, neuroglancer 2.22, python 3.8.5 Chromium 90.0.4430.212 Fedora Project, ChromeDriver 90.0.4430.212 , served by cors_webserver.py

We saw it in the script “ Frames are captured only once all data has been loaded; because of
this, rendering can be slow.” But we couldn't understand why it was so slow, especially since the amount of data didn't feel very large.