Five records · Mare Vitalis (2000) · Low Level Owl I & II (2001) · Two Conversations (2003) · Illumination Ritual (2013)
Every original track, each pointed at the best Songsterr destination for it and labeled with tuning, instruments, drums, and whether a person made the tab or Songsterr auto-generated it.
56 original tracks10 on Songsterr4 AI primaries6 human primaries1 have human + AI10/56 with a tab46 tab-less tracks linked to the song on YouTube9 now in your own account
How the primary link is chosen. Where both a human and an AI tab exist, the AI one is the primary link, since the AI transcriptions here are usually the more complete. The human tab stays as a labeled alternate. A live-only AI tab is not promoted over a studio human tab. An HTTP 200 proves a link responds, not that its notation was checked against the record.
AI-generated tabs. Songsterr's aiGenerated flag marks each tab human-made or AI auto-generated. 1 songs have both a human and an AI tab; the AI is primary except a live-only AI take, which stays an alternate. Counts are computed from the audit file. Songsterr splits this band across two ids: a189357 (The Appleseed Cast) and the older a665 (Appleseed Cast). Both are checked and keyed by exact songId.
Nine of these are now in your own account, and six hold only drums. Songsterr's AI writes every instrument from a YouTube link, and its importRevisionInput replaces the whole tab rather than one track. Importing a drums-only .gp5 over an AI transcription therefore deletes the guitar and bass it wrote. That happened to six songs on 6 August 2026. The three marked YOURS still carry all their instruments. The right order is to download the AI file, swap only the drum track, and import that. Every tab here is private to your account. Track counts read from /api/meta/{songId}/revisions on 7 August 2026.
AI Songsterr auto-generatedH+AI a human and an AI tab both existDRUMS has a drum part · NO DRUMS noneNO TAB nothing on Songsterr · LISTEN the song itself on YouTubeYOURS a tab in Brandon's account with every instrument · YOURS the corrected drum part alone
All parts, rebuilt from your stems
Songsterr keeps one revision per song, so the AI guitar and bass that the drums-only import replaced on 6 August 2026 are not recoverable from the site. These files are built the other way: guitar and bass transcribed from the six-stem split, merged onto the corrected stem-built drum track, on that track's own tempo grid.
How the transcription was checked. Mare Mortis is the only song whose real Songsterr guitar and bass were saved to disk before the overwrite, so the stem-built parts were scored against it bar by bar. Guitar reached 97.3% bar coverage and 0.889 pitch-class recall; bass reached 97.1% and 0.672. Recall peaks at zero semitones of shift (guitar 0.889 against 0.388 for the next best shift), which is what establishes the grid and the key rather than a lucky near-miss. A second check ran on a song nothing was tuned against: On Reflection, whose Songsterr tab still holds all five tracks. Over bars 1 to 26, before that tab changes from 120 to 135 BPM, the stem-built guitar scores 25 bars matched, 0 missed, 0.986 pitch-class recall against the reference Lead and Rhythm parts combined.
One bug caught before any of this shipped. The song length was being taken from the drum track, and drums stop before songs do. On The Last in a Line the drums end at 0:40 of a 4:13 song, so the tab ended at bar 22 and took the guitar and bass down with it, cutting 216 seconds of playing out of the file. It hit 13 of 32 files. The grid now spans the audio and the drums pad with rests, which took that song from 22 bars and 219 guitar notes to 152 bars and 2,285. Every file here now covers 100% of its own audio, measured by re-parsing each one.
Six files were carrying notation over stems that hold no signal, and that is now fixed. A note count cannot tell a sparse song from an invented one. This record is full of ambient interludes, so a thin drum tab is often correct, and the only way to separate the two is to measure the stem the part claims to transcribe. Every lane is now checked for level, and a part over a dead lane is emptied rather than written. Messenger was the worst of it: an interlude with no drums at all, whose kit measures -87.2 dBFS at a peak of 0.001, carrying 319 written drum notes that onset detection had produced from digital silence. The other five were bass parts built from bleed in a silent bass stem.
Song
Part
Notes removed
Stem level
Peak
Messenger
Drums
319
-87.2 dBFS
0.001
Bird of Paradise
Bass
522
-63.5 dBFS
0.002
Convict
Bass
348
-71.9 dBFS
0.003
The Immortal Soul of Mundo Cani
Bass
194
-82.1 dBFS
0.001
A Tree for Trials
Bass
115
-72.0 dBFS
0.001
The Waking of Pertelotte
Bass
81
-75.8 dBFS
0.006
The check was held to controls in both directions before it was trusted. Sentence has a real but sparse kit at -47.9 dBFS with a 0.770 peak, and it keeps its hits with only its dead hat and tom lanes dropped. On Reflection has a full kit and loses nothing, staying at 2,736 detected hits. A part marked silent in the table above is empty on purpose, because that stem has nothing on it to transcribe.
Two limits to note. A six-stem split gives one guitar stem, so lead and rhythm arrive as a single Guitar track; separating them needs the stereo-field pass before demucs. And the drum tempos carry real uncertainty: the tempo recheck left most of this album at OCTAVE_CONFLICT or LOW confidence. Fitting drum onsets to a grid favours 131.75 BPM on Mare Mortis where the Songsterr tab says 125. Guitar, bass and drums are generated onto one shared grid, so they agree with each other at any tempo; a wrong tempo changes speed against the record.
Tunings, capo and delay, checked against outside sources
Every one of the 46 files on this page was written in a single tuning, Drop D on guitar and bass. That is wrong for at least four songs, and the catalogue plainly uses more than one tuning. What follows is what free sources actually state, with each row labelled by where it came from.
Song
Source
Author
Independent
Guitar tuning
Bass tuning
Capo
Fishing the Sky (Mare Vitalis 2)
Songsterr songId 1580
Darr
yes
E A D G B E Standard EADGBE
E A D G Standard EADG
none stated
Forever Longing the Golden Sunsets (Mare Vitalis 3)
Songsterr songId 2038
Leland
yes
E A D G B E Standard EADGBE
E A D G Standard EADG
none stated
Steps and Numbers (Low Level Owl I 7)
guitaretab via chordie 281434, and Ultimate-Guitar tab 1067048
community
yes
E A D G B E Standard EADGBe
not given
none stated on the tab
30 Degrees 3 Am (Illumination Ritual 5)
Songsterr songId 1125062
John Howard (AI)
yes
E A D G B E Standard EADGBE
E A D G Standard EADG
none stated
On Reflection (Low Level Owl I 2)
Songsterr songId 6206980
Brandon chavez (AI)
no, Brandon's own upload
D# G# C# F# A# D# Eb standard, half step down
D# G# C# F# Eb bass, half step down
none stated
Hanging Marionette
Songsterr songId 1125058
John Howard (AI)
yes
F A C G C E F A C G C E, alternate
F A C G F A C G, alternate
none stated
Moment
Songsterr songId 4063963
Richard Sanchez (AI)
yes
D A D G B E Drop D
D A D G Drop D bass
none stated
Novice
Songsterr songId 3849400
Savarveribuk (AI)
yes
D A D G B E Drop D
D A D G Drop D bass
none stated
On Sidewalks
Songsterr songId 3721562
Savarveribuk (AI)
yes
D A D G B E Drop D
D A D G Drop D bass
none stated
Four songs on this page are demonstrably in the wrong tuning. Independent tabs give Standard EADGBE for Fishing the Sky, Forever Longing the Golden Sunsets, Steps and Numbers and 30 Degrees 3 Am, while the files here declare Drop D. Two of those four sources are human-written rather than machine-generated, which is the strongest evidence available for free. The sounding pitches in these tabs are still the ones measured off the stems; it is the fret positions that are wrong, because they were computed against a low D that the record does not have. Retuning them means regenerating the fingerings, not editing a tuning line, so these files are unchanged until that rebuild runs.
What the recordings themselves settle, and what they cannot. Measuring how far the strong spectral peaks miss the equal-tempered grid gives a guitar median of +4.3 cents across 46 songs. That says these records sit at concert pitch with no tape-speed drift, which is what makes the transcription grid trustworthy. It cannot tell Standard from Eb from Drop D, because all three still play notes on the same A440 grid, so it is not used here as tuning evidence. The lowest note each guitar lane genuinely sustains is more useful, and it is not one value: D2 in 22 songs, D#2 in 7 songs, E2 in 14 songs. Three different floors across one catalogue is by itself proof that a single tuning cannot be right for all 46.
Capo: nothing found, and nothing invented. No capo position is documented for any Appleseed Cast song in any free source checked.
Songsterr's track metadata has no capo field at all; the schema is difficulty, hash, instrument, instrumentId, isEmpty, isVocalTrack, name, tuning, views.
The Steps and Numbers tab states a tuning and explicitly carries no capo line.
Christopher Crisci does not mention a capo in the Low Level Owl 10-year retrospective interview.
Christopher Crisci does not mention a capo in the Greg's Guitar Lessons interview.
Left blank rather than guessed. The measured lowest sounding note per song is published instead, since that is the objective number a player can check a capo against.
Delay
Delay is central to the guitar sound and is documented per player, across two eras.
Player
Era and source
Delay
Rest of the board
Christopher Crisci
effectsbay.com, November 2011, pedalboard feature
Boss DD-7 (possibly DD-6), Line 6 DL4 Delay Modeler, Electro-Harmonix Deluxe Memory Man, older vintage version
Boss TU-2 Tuner, Boss TR-2 Tremolo, Boss RC-20XL Loop Station, Spaceman Aphelion Overdrive, Fulltone Fulldrive2 MOSFET
Christopher Crisci
gregsguitarlessons.com, 2 July 2011
Line 6 DL4, Boss DD-5
Christopher Crisci
equipboard.com artist page
Caroline Guitar Company Kilobyte Lo-Fi Delay, EarthQuaker Devices Avalanche Run, Electro-Harmonix XO Deluxe Memory Man
Tech21 SansAmp Bass Driver DI, Electro-Harmonix Cathedral Stereo Reverb, Boss ODB-3 Bass Overdrive, Korg Pitchblack Tuner | No delay on the bass board at all, which matters for reading the bass tabs: repeats in the low end are the room and the guitars, not the bass rig.
No source checked states delay times, feedback or mix values. None are published here. Delay times, feedback and mix are not published by any source checked, so none are stated here.
Every file above carries a single Guitar track, because the build transcribes the mono sum of the guitar stem. The reason on record for that was that a six-stem split returns a mono guitar. That is not true of these recordings. All 46 guitar stems are two-channel, and on 26 of them the two channels are decorrelated enough to be holding different parts.
Checked rather than assumed. Correlation alone could be reverb width, so Steps and Numbers (correlation -0.009, side energy level with mid) was split and each channel transcribed on its own. The left channel returns 2,562 notes, the right 2,980, and only 29.2% of left onsets have a right onset at the same pitch within 50 ms. One part double-tracked would agree far higher than that. The pitch-class profiles differ too: F# is 12.5% of the left and 17.5% of the right, A# is 4.4% left against 1.6% right. These are two players, not one take widened.
The four-track file, built and downloadable. Guitar 1 (left) 2,549 notes, Guitar 2 (right) 2,759, Bass 896, Drums 1,912, all 192 bars at 130 BPM. Bass and drums are identical to the three-track version, so nothing was disturbed. The summed read gave one Guitar track of 3,068 notes, meaning the mono sum was masking roughly 2,240 notes of real playing. Steps and Numbers, 4 tracks .gp5 ↓
Song
L/R correlation
Side minus mid
Reading
Low-Level-Owl-II-02-Strings
-0.023
+0.2 dB
two parts, separable
Low-Level-Owl-I-07-Steps and Numbers
0.003
-0.0 dB
two parts, separable
Mare-Vitalis-07-...And Nothing Less
0.005
-0.0 dB
two parts, separable
Mare-Vitalis-01-The Immortal Soul of Mundo Cani
0.007
-0.1 dB
two parts, separable
Mare-Vitalis-09-Kilgore Trout
0.007
-0.1 dB
two parts, separable
Mare-Vitalis-03-Forever Longing the Golden Sunsets
0.022
-0.2 dB
two parts, separable
Low-Level-Owl-II-01-View of a Burning City reprise
0.048
-0.4 dB
two parts, separable
Low-Level-Owl-I-06-Doors Lead to Questions
0.056
-0.3 dB
two parts, separable
Mare-Vitalis-02-Fishing the Sky
0.067
-0.6 dB
two parts, separable
Low-Level-Owl-II-03-A Place in Line
0.069
-0.6 dB
two parts, separable
Mare-Vitalis-04-Mare Mortis
0.100
-0.8 dB
two parts, separable
Mare-Vitalis-06-Secret
0.103
-0.9 dB
two parts, separable
Low-Level-Owl-I-11-Convict
0.124
-1.1 dB
two parts, separable
Mare-Vitalis-05-Santa Maria
0.131
-1.1 dB
two parts, separable
Low-Level-Owl-I-02-On Reflection
0.139
-1.2 dB
two parts, separable
Low-Level-Owl-I-14-View of a Burning City
0.153
-1.3 dB
two parts, separable
Low-Level-Owl-II-11-Reaction
0.204
-1.8 dB
two parts, separable
Low-Level-Owl-I-10-Mile Marker
0.261
-2.3 dB
two parts, separable
Low-Level-Owl-II-10-The Argument
0.268
-2.3 dB
two parts, separable
Low-Level-Owl-I-05-Messenger
0.273
-2.4 dB
two parts, separable
Low-Level-Owl-I-12-A Tree for Trials
0.278
-2.4 dB
two parts, separable
Low-Level-Owl-II-04-Shaking Hands
0.289
-2.6 dB
two parts, separable
Low-Level-Owl-II-07-Sunset Drama King
0.291
-2.6 dB
two parts, separable
Low-Level-Owl-II-05-Rooms and Gardens
0.326
-2.9 dB
two parts, separable
Low-Level-Owl-II-06-Ring Out the Warning Bell
0.336
-3.0 dB
two parts, separable
Low-Level-Owl-I-03-Blind Mans Arrow
0.365
-3.3 dB
two parts, separable
Low-Level-Owl-I-13-Signal
0.453
-4.2 dB
one part, double tracked
Mare-Vitalis-10-Storms
0.459
-4.3 dB
one part, double tracked
Illumination-Ritual-07-Barrier Islands (Do We Remain)
0.497
-4.7 dB
one part, double tracked
Low-Level-Owl-I-01-The Waking of Pertelotte
0.499
-4.3 dB
one part, double tracked
Mare-Vitalis-08-Poseidon
0.508
-4.6 dB
one part, double tracked
Illumination-Ritual-01-Adriatic to Black Sea
0.563
-5.5 dB
one part, double tracked
Low-Level-Owl-II-09-Decline
0.569
-5.6 dB
one part, double tracked
Low-Level-Owl-II-08-The Last in a Line
0.598
-4.2 dB
one part, double tracked
Illumination-Ritual-03-Simple Forms
0.601
-6.0 dB
one part, double tracked
Low-Level-Owl-I-09-Bird of Paradise
0.607
-5.2 dB
one part, double tracked
Illumination-Ritual-10-Illumination Ritual
0.609
-5.9 dB
one part, double tracked
Low-Level-Owl-I-04-Flowers Falling From Dying Hands
0.708
-7.7 dB
one part, double tracked
Illumination-Ritual-09-Clearing Life
0.719
-7.5 dB
one part, double tracked
Illumination-Ritual-08-North Star Ordination
0.757
-8.5 dB
one part, double tracked
Illumination-Ritual-02-Great Lake Derelict
0.790
-9.2 dB
one part, double tracked
Illumination-Ritual-04-Cathedral Rings
0.791
-9.1 dB
one part, double tracked
Low-Level-Owl-I-08-Sentence
0.811
-9.7 dB
one part, double tracked
Low-Level-Owl-II-12-Confession
0.876
-11.7 dB
one part, double tracked
Illumination-Ritual-05-30 Degrees 3 Am
0.899
-12.6 dB
one part, double tracked
Illumination-Ritual-06-Branches on the Arrow Peak Revelation
0.901
-12.8 dB
one part, double tracked
20 songs stay at one guitar on purpose. Their channels correlate between 0.4 and 0.9, which is the signature of one part recorded twice rather than two parts. Illumination Ritual is nearly all of these. Writing a second guitar there would invent a part instead of recovering one. Not one song in the catalogue correlates above 0.95, so a genuinely centre-summed guitar never existed here.
Two guitars, and a hand can play both
Two guitar tracks per song, Guitar 1 from the left channel and Guitar 2 from the right, which is the split the stereo probe already established. The fingerings that shipped before were placed by a rule comparing each note against the single note before it, with a 24-fret ceiling and no idea where the hand was, and that produced 1111 chords wider than any reach and 867 notes above the 19th fret. Every note has been re-placed as a shortest path through hand positions. No chord on any staff is now wider than a four-fret hand span, on any of the eight songs.
What the gate proves. Pitch is recomputed from string, fret and tuning on both sides of the change and the ordered per-beat sequences are compared, so a refret that quietly moved a note fails. Every guitar track reports 0 pitch changes. Notes sitting above the top of the instrument were folded by whole octaves rather than deleted, and each one is counted. Every drum, bass and vocal track is untouched, and the AI's vocal line on A Tree For Trials verifies identical at 39 note events and 120 lyric characters.
What it took, and what it cost. Three approaches were measured on Blind Man's Arrow before this one. Refretting alone took 189 unreachable chords to 17. Deleting quiet notes until each staff matched the density of Songsterr's own AI reached 2, and threw away 58% of the playing to get there. Splitting each guitar into a low and a high staff reached 8 and kept everything, but it needs four staves and the record has two guitars. What shipped keeps two staves and reaches 0: the fingering is solved first, and only a beat that still cannot be held loses notes, furthest-from-centre first, one at a time, stopping the moment the rest fit. Across all eight songs that is 26813 of 28391 notes kept, and the removals are only ever notes that were physically impossible to fret together.
All eight are live on Songsterr, five tracks each, read back off CloudFront rather than off the editor screen: Guitar 1, Guitar 2, bass, drums and the vocal line. On Reflection 8421499, ...And Nothing Less 8421544, Blind Man's Arrow 8421353, Mile Marker 8421388, Secret 8421427, Kilgore Trout 8421459, A Tree For Trials 8422068, Mundo Cani 8422125. Six of the eight had never been published at all. Blind Man's Arrow verifies live at 0 unreachable chords on both guitars with a worst reach of 4 frets.
The last two needed a bass that demucs had hidden. A Tree For Trials and Mundo Cani were held back because their AI tabs carried no bass, and the stem pipeline had none to give: both `bass.wav` lanes are empty. Measuring 40-160 Hz per lane found the low end sitting in the GUITAR stem, roughly 60 dB above the lane named for it (A Tree For Trials, bass -21.5 dB against guitar 43.8; Mundo Cani, -17.9 against 38.6). That also corrects the earlier note in STATE.md, which recorded these as piano-misroutes. Low-passing the guitar stem to the bass band and tracking it with crepe, which is monophonic and so does not report the harmonic series as extra voices, recovered 59 and 47 notes. A Tree For Trials also got its drums, 325 hits written straight into the GPIF from the stem-built drum file.
A Tree For Trials had one staff for two guitars. Songsterr's AI wrote it as a two-track tab, and the record has two players: the isolated left and right channels correlate at 0.28, which is two performances rather than one panned wide. A second staff was added and fed from the left channel, the first stays on the right, and the tab went from 297 guitar notes on one staff to 661 across two. Its grid also stopped at bar 24, covering 68.6s of an 88.9s record, so 26 notes were being dropped on the way in; it now runs 31 bars and 88.6s, inside the 2% the import gate allows.
0 chords still cannot be played by one hand, and that is the honest answer. They are beats where the stem holds two guitars at once, so the notes are genuinely simultaneous and no fingering exists: on A Tree For Trials, bar 3 asks for B2, G4 and G5 together, which needs the 2nd fret and the 15th at the same moment. Where the two players are separable by channel they now get separate staves. Where one channel still carries bleed from the other, the beat is notated at the narrowest reach that holds the right pitches and left visible rather than quietly thinned.
Record 1
Mare Vitalis
2000 · Deep Elm4 human · 0 AI · 6 unavailable
1
The Immortal Soul of Mundo Cani YOURS · 4 trkNO TABLISTEN
No Songsterr tab under any checked profile. Video: The Appleseed Cast - Topic.
Your tab s6285808: 4 tracks, all instruments, unpublished and private to you.