Video-by-code tools, actually run and measured.

videobycode tests tools that generate video from code, and publishes what the runs actually produced

Comparisons, render costs, and working implementations for tools that generate video from code. Every number here comes from a run we performed, with tool version, operating system, and wall-clock render time recorded.

How we test

What a recorded run looks like

How we test
6s excerpt from 30s output
Tool
ffmpeg 8.0.1 (libx264, aac; libfreetype=no)
Wall clock
7,549 ms
Output
1920×1080, 30 fps, 989,397 bytes

Runvbc-2026-08-09-ffmpeg-tpl30-001, measured 2026-08-09. Excerpt starts at 2s of the TPL-30 output; the clip is re-encoded to VP9 for the web and carries no audio track. Speed and colour are unchanged.

What we publish

Recorded on every test

Editorial policy

Tool and version

The exact release we installed, so a result can be checked against the same build later.

Machine and OS

CPU, memory, and operating system. Render numbers mean nothing without the machine they came from.

Wall-clock render time

Measured end to end on the stated output settings, not the figure printed by the tool.

Attempts, including failures

How many runs it took and what broke on the way. Runs that never finished are reported too.

Article types

How we test

Measured

Numbers produced by running the tool ourselves, with the setup written down.

Tutorial

A task implemented step by step, with the code and the run time included.

Comparison

Tools scored on one fixed set of criteria, stated before the results.