IMF playback speed is too fast compared to in-game/other players

[snoopin1]

The playback rate of IMF files is faster than it is in-game on real hardware and in other players.

I have attached an image comparing different recordings of the end of the kickpant IMF (560 Hz playback rate) from Commander Keen 4, with each recording synced up at the start of the song (or nearly so). Note how relatively far the AdPlug/Winamp version has drifted compared to the others after just 32 seconds. I confirmed that this happens both with real-time playback and when using Nullsoft's Disk Writer plugin. K1n9_Duk3's IMF Player and Lameguy64's PLAYIMF both play the IMF in DOSBox Staging at nearly the same speed as the game/foo_input_adplug, which I believe is further evidence that AdPlug's IMF playback code is inaccurate.

I have tested other IMFs as well, such as the 280 Hz ones from Duke Nukem II, and they play faster in AdPlug/Winamp as well.

foo_input_adplug.wav

[mywave82]

Can you attach what you expect to be the correct audio to this bugreport, and we can compare it to the latest version (including adplugdb)

[snoopin1]

I have attached a WAVE of foo_input_adplug's output of the IMF to the original post. I believe the speed of the song in this recording is accurate, as imf2vgm's timing logic produces a VGM that plays at the same speed with VGM Input Plugin.

[Malvineous]

It looks like the difference is very small, a few hundred milliseconds over the length of the song?

I wonder if it's down to the recommended delays programming the hardware. After writing to the OPL chip you're supposed to wait for a certain number of microseconds before sending the next byte. The emulated chips don't have this limitation so effectively writes are instant. Perhaps this lack of microsecond-level delays adds up over the length of the song?

[dynamite1981]

You could use adplugdb to adjust the playback speed for IMF files that sound off. The default is 560.44 Hz for KICKPANT.IMF. Perhaps it should instead be 560 Hz precisely?

…

On Wed, Mar 18, 2026 at 4:42 AM Adam Nielsen ***@***.***> wrote: *Malvineous* left a comment (adplug/winamp#31) <#31 (comment)> It looks like the difference is very small, a few hundred milliseconds over the length of the song? I wonder if it's down to the recommended delays programming the hardware. After writing to the OPL chip you're supposed to wait for a certain number of microseconds before sending the next byte. The emulated chips don't have this limitation so effectively writes are instant. Perhaps this lack of microsecond-level delays adds up over the length of the song? — Reply to this email directly, view it on GitHub <#31 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADDG2X2G2GOYRY33YIACVCD4RKDSHAVCNFSM6AAAAACWUYIKZ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DAOBRHA2DGNBZGI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[snoopin1]

You could use adplugdb to adjust the playback speed for IMF files that
sound off. The default is 560.44 Hz for KICKPANT.IMF. Perhaps it should
instead be 560 Hz precisely?
…

Perhaps, but the issue affects all IMFs regardless of what their playback rates are set to. The Duke Nukem II IMFs for example are defined in the database as having a 280 Hz playback rate, but AdPlug still plays them slightly faster than normal. "He's Back!" drifts by over 300 milliseconds towards the end of the file. Although this difference is very small, it's enough to subtly affect the sound of instrument envelopes at certain points in the music.