Draw a spectrogram picture of a sound file with the ffmpeg showspectrumpic filter on Windows,…

Guide ID: art_01M7DJ5K10A91HZMZKHF7PRMA5

Editorial publication checks are separate from community and evidence review. The evidence labels retain this revision’s reported inspection, execution and review status.

Guide title

Draw a spectrogram picture of a sound file with the ffmpeg showspectrumpic filter on Windows, and check it: a script that writes the PNG, reads it back as pixels and tells you where the tones you name were drawn, in pixels and in Hz

Coverage

A PowerShell script that draws a spectrogram PNG of a sound file with the ffmpeg showspectrumpic filter, reads the picture back as pixels and reports where the tones you name were drawn, in pixels and in Hz, with its tolerance. Generated test tones show what the manual leaves out: the legend is on by default and makes the picture larger, channel one is drawn at the bottom, the log axis misplaces low tones, a quiet tone vanishes with exit code 0 unless the dynamic range is a few decibels above its depth, a 16-bit file can leave a faint false line, a pass cannot tell an error smaller than the tolerance, and ffmpeg 9.0.2 sometimes crashes on a six-channel picture. No real recording was used and nothing was looked at by eye.

Example

On a generated mono file with tones at 1000, 5000 and 10000 Hz, a request for 1024 by 512 came out 1308 by 640 with the legend on, and all three lines were within one pixel of the expected places. A tone at minus 60 dBFS stayed all black at a dynamic range of 64 and showed from 65. Nothing was looked at by eye.

Useful starting points

• Print the plan first: with the legend on the picture is larger than the size you ask for.
• Read the offset in Hz as well as in pixels: at the default size the tolerance is 94 Hz, so a small frequency error can still pass; a taller size gives a finer check.
• For several channels use the separate mode and remember that channel one is at the bottom.
• If a quiet signal does not show, give the dynamic range several decibels more than the depth of the signal, or raise the gain; a tone at the edge is all black without any error.
• Open the PNG yourself: the script checks positions and brightness, not whether the picture is useful.