Skip to content
TVGEP

Tube Slider

North American Tube Slider GameCube front cover, showing a red racing vehicle inside a tube.
Box art source · North America

Nintendo GameCube · 2003

Tube Slider

Developer
NDcube
Region
USA
Disc ID
GTUE8G
Release information source

Tube Slider is a futuristic racing game developed by NDcube and published by NEC Interchannel for the Nintendo GameCube in 2003.

I played this game a lot growing up. It sat comfortably alongside the likes of other racing titles at the time, such as Need for Speed Underground 2, Hot Pursuit, and F-Zero GX. I’m still surprised at the lack of discussion about this title online. This game earned its entry after I began extracting the OST from the game’s files and found a suspicious-looking debug flag.

Confirmed in-game · Verified one-byte patch

The debug menu

Link to this discovery ↗

Redirecting one script entry makes the examined game image boot into its shipped debug menu. The patched image was tested successfully in Dolphin.

Tube Slider debug menu with Normal Start, Memory Card Utility, and Sound Test Menu entries.
Main debug menu, reached using the one-byte entry patch in Dolphin. Original PNG
Tube Slider sound-test screen showing SE No. 0000 and BGM No. 0000.
Sound-test screen from the patched image, with BGM set to 0000. Original PNG

What remains unknown: Access through unmodified retail gameplay has not been established.

Read about how this was found: The debug menu

Recovering the startup entry

Section titled “Recovering the startup entry”

The startup code loads mac/system.mac and selects entry 0. The script header contains two entry offsets:

Header offset 0x08: 00 00 00 1C   entry 0: normal startup
Header offset 0x0C: 00 00 00 5A   entry 1: debug startup

Changing the byte at file offset 0x0B from 1C to 5A redirects entry 0 to the existing debug entry. These are offsets within the extracted macro file, not memory addresses. The executable is unchanged.

The examined native startup sets r6 = 0 at 0x80012C64, then calls the macro loader at 0x80012C74. The loader indexes the entry table at header offset 8 + 4 × r6. The alternate script entry invokes the debug state dispatcher.

Observed controls

Section titled “Observed controls”

The patched image boots into the debug menu in Dolphin. Testing confirmed these controls:

  • Analog stick: navigate.
  • Left / right: change line values.
  • A: confirm or advance into a menu.

The captures above show the debug menu and sound-test screen from the patched image. They are unmodified Dolphin PNG captures taken after selecting 1× native resolution, Auto aspect ratio, and disabling the widescreen hack and visual enhancements. They are displayed proportionally. An instrumented runtime trace is not included; these still images do not establish Track 13 playback.

Exact image used

Section titled “Exact image used”

The original image is a 664,830,528-byte CISO of the USA release. A full original-versus-patched comparison found exactly one differing byte.

Location Offset
Extracted mac/system.mac 0x0B
Logical disc 1196017747 (decimal)
Physical CISO 399132755 (decimal)

Original CISO SHA-256

d178a9e25cf285e82268cecae8da6b6f736146dd0ea4c7c84e3e30e4c63be0cf

Patched CISO SHA-256

2703374a44ac1ce18c6cc85e62bf269ab3d282dae3f67fb37ecb0607cc1f771d

To reproduce on the identical source image, first verify its hash, work on a copy, change the byte at the physical CISO offset above, then verify the patched hash. Physical container offsets are specific to this image layout; they must not be applied blindly to ISO, RVZ, or another CISO.

Measured audio difference · Static playback reference

An alternate music track

Link to this discovery ↗

Track 13 has stronger bass relative to treble than Track 10. The recovered debug sound-test script contains a branch that selects Track 13 for playback.

Compare Tracks 10 and 13

Matching excerpts from 0:30–1:00. Original levels preserved: Track 13 is louder. No normalization, EQ, fades, or resampling applied.

Open Track 10 WAV
Open Track 13 WAV

Excerpt provenance and SHA-256 hashes

What remains unknown: Playback of Track 13 in the sound test has not been runtime-confirmed. Normal gameplay use remains unresolved. The measurements do not prove that EQ was the only processing applied.

Read about how this was found: An alternate music track

Tracing the playback reference

Section titled “Tracing the playback reference”

The debug menu contains a SOUND TEST MENU entry. Its BGM selection subroutine has branches for tracks 0 through 13. The branch for 13 contains:

system.mac offset 0x0CF6: 80 02 08 1F 00 00 00 0D
system.mac offset 0x0CFE: 00 C4 08 00 00 0E BF 1F
system.mac offset 0x0EBF: snd/BGM13_S.ADP\0

The first instruction tests the BGM register against 13. Opcode 0xC4 in the second instruction resolves the filename at 0xEBF. Its native handler at 0x80011490 calls the same filename-taking audio routine (0x80074F7C) used by other BGM paths. This is a playback reference, rather than just a filename sitting in the script.

A literal filename scan of the executable and all 270 disc file payloads found Track 13 names only in system.mac. The inspected native numbered-BGM table contains tracks 00–10 and an empty entry. These findings narrow the known uses; they do not exclude indirect or constructed references elsewhere.

Comparing the decoded audio

Section titled “Comparing the decoded audio”

Listening identified Tracks 10 and 13 as versions of the main-menu music. The decoded files have identical lengths: 11,296,768 frames, or about 235.35 seconds at the decoder’s 48 kHz output rate.

The measurements support a common underlying recording with altered frequency balance. Track 13 is 4.10 dB higher in overall RMS, with a larger increase in bass than treble:

Frequency band Track 13 minus Track 10 After equalizing overall RMS
20–60 Hz +6.45 dB +2.35 dB
60–120 Hz +5.02 dB +0.92 dB
120–250 Hz +4.30 dB +0.20 dB
1–2 kHz +1.99 dB −2.11 dB
4–8 kHz +1.37 dB −2.73 dB

Frequency analysis of Tracks 10 and 13, comparing their spectral power and frequency-dependent level difference.

The deepest bass is therefore about 5.08 dB stronger relative to the 4–8 kHz band in Track 13. Energy-weighted coherence is approximately 0.9715, consistent with closely related source audio. It does not prove that EQ was the only processing applied or recover the original producer’s settings.

Measurement provenance

Section titled “Measurement provenance”

Inputs were snd/BGM10_S.ADP and snd/BGM13_S.ADP, decoded using vgmstream r2117 as Nintendo DTK ADPCM, with loops ignored. Analysis used the complete decoded tracks: per-channel Welch power spectra and cross-spectra, a 16,384-sample Hann window, 50% overlap, and summed stereo power. Equal-RMS results remove the overall level difference; they are not perceptual loudness normalization. The 48 kHz rate describes the decoded output, not an independently measured in-game playback rate.

Download the full frequency measurements (JSON).