Published Notation vs the Songsterr GPs

Over-Nite Sensation & Roxy & Elsewhere · pass 1 Zomby Woof · pass 2 Penguin In Bondage · pass 3 Village Of The Sun · pass 4 Cheepnis
17songs matched
92notation images
43Songsterr tabs
6published signature labels
1 of 3tabs use all six labels
4songs compared

The question

Songsterr's Zappa tabs are community and AI made. The archive built on 28 August holds published notation for these same two records, so each tab can be checked against a source with real editorial standing rather than taken on trust. One song per pass. This is pass 1, Zomby Woof, chosen because it has two independent published sources, three tabs to compare, and the odd-meter writing that makes a transcription easy to get wrong.

What the published side actually says

Two independent sources. Their overlap is limited, so what they give is combined coverage without contradiction rather than mutual corroboration. Sloots documents one passage, Brown documents a different one, and neither confirms the other's bars.

1. Kasper Sloots / Paul Pappas, zappa-analysis.com

Full score of Zomby Woof from 1:39 to 2:10, updated fall 2024, deposited at the I-depot in The Hague. He identifies it as a transcription from the recording. A separate interlude image on the same page is reproduced from the Over-Nite Sensation Guitar Book, page 64, serving as a comparison rather than as the source of the excerpt. He prints thirteen bars, running 7/8, 5/4, 10/16 ×4, 7/8, 6/4, 4/4 ×5.

2. Ryan Brown, DRUM Magazine

"It has lots of fives, including groups of five sixteenth-notes and quintuplets. The song also has a lot of different time signatures."

"I always hear the 15/16 section as two groups of notes: one group of four eighth-notes, plus one group of seven sixteenth-notes."

Four eighths is 8/16, plus seven sixteenths is 15/16. The arithmetic checks.

Combined published signature labels for Zomby Woof: 4/4, 5/4, 6/4, 7/8, 10/16, 15/16. Six labels drawn from two sources covering different passages. No label here is confirmed by both.

The three Songsterr tabs against that set

TabBarsBPMDurationDrum notesParen. Published meters presentMeters written
Community, hand-tabbed
s412162 · by meh32 / Ben Dibden1
114885:03.861,574326 of 63/4, 10/16, 2/4, 5/4, 15/16, 4/4, 6/4, 7/8
Incomplete upload
s5820647 · AI run by CoolDude
123895:30.84182102 of 613/16, 4/4, 9/8, 7/8
Yours
s6685613 · your own AI run, private
2321755:08.232,06303 of 64/4, 5/4, 2/4, 6/4, 5/8, 3/4

Album track length is 5:11. Drum-track data read from Songsterr's own revision API.

One of the three tabs uses all six published signature labels. The community hand tab carries every one, and its bars 38 to 50 run 7/8, 5/4, 10/16 ×4, 7/8, 6/4, 4/4 ×5, matching the thirteen bars Sloots prints, in order. That makes it the strongest provisional bar map. A label matching a label is not yet a verified transcription.
The incomplete upload flattens the song. 119 of its 123 bars are 4/4. It has none of 5/4, 10/16 or 15/16, and it carries a 13/16 and a 9/8 that neither published source shows. Its drum staff holds 182 real notes against the hand tab's 1,574.
Correction, 29 August. A first version of this page said this tab carried "297 notes, 115 of them null-pitched, so the part is malformed". That was wrong. Songsterr's JSON emits a {"rest": true} object inside each beat's note list, and I counted those rests as notes. The 115 are rests, which is ordinary notation. Every note count on this page is now rest-excluded.
Yours is the densest drum candidate with the closest calculated duration. 2,076 drum notes shows density. 5:08.23 against the album's 5:11 shows approximate structural coverage. Zero parenthesised notes shows compliance with the house file rule. None of those three measurements establishes accurate drums, appropriate cymbals, playable orchestration, or correct attack placement, so it stays a detailed repair candidate pending comparison against the audio. It writes at 175 bpm in 232 bars, double-time against 88 bpm in 114, a legitimate choice. What it does not carry is the 7/8, 10/16 or 15/16 labelling, with twelve bars of 5/8 in their place.
Provenance, corrected 29 August. These three are not "hand versus AI versus yours". Songsterr's API records s412162 as editor-made by meh32 over Ben Dibden1, and BOTH s5820647 and your s6685613 as AI runs, the first uploaded by CoolDude and the second your own. What separates them is how completely each was uploaded. Measured per staff, rests excluded:
Staffs5820647 (CoolDude)s6685613 (yours)
Lead Guitar0651
Rhythm Guitar108532
Bass1211,176
Drums1822,063
Vocals0380
Total4114,802

Two of the five staves on s5820647 contain rests and nothing else. A survey of 34 Zappa tabs across both records found this on three of them, and never on a tab built in the editor. Incomplete uploads are the pattern here, and they appear across tools.

The alignment test

A matching list of labels proves less than it looks like, since the same labels could sit in the wrong bars. Sloots states his own timestamps and prints thirteen bars, so the tab's tempo map gives a check on both the span and its contents.

Span mappedBarsTab durationvs the 31.0 s excerpt
b38 to b46921.33 s−31.2%
b38 to b481126.67 s−14.0%
b38 to b501332.00 s+3.2%
b38 to b521537.33 s+20.4%
Correction, 29 August. A first version of this page mapped nine bars and reported the alignment as 2 percent. That figure came from dividing each endpoint offset by its absolute position in the song, which measures nothing about alignment. Across the excerpt itself the nine-bar span is 31.2 percent short. The fault was mapping nine bars against the thirteen Sloots prints.

Thirteen bars is both the count Sloots prints and the closest duration match, at 32.00 s against 31.0 s. Its endpoints sit at +2.07 s and +3.07 s, so the tab runs about one second long across the passage. The thirteen-bar run is 7/8, 5/4, 10/16, 10/16, 10/16, 10/16, 7/8, 6/4, 4/4, 4/4, 4/4, 4/4, 4/4, which is the order and the page breaks Sloots prints.

Computed from the tab's own signatures and its six tempo marks (88, 76, 88, 90, 82, 88 bpm). This corroborates the bar map at the level of labels and span. It is not a per-bar proof of the notes, and confirming that needs the album master.

Verdict for Zomby Woof

Use it forTakeWhat the measurement supports
Bar map and signature labelsCommunity s412162 The strongest provisional bar map. All six published labels, in the printed order, across a thirteen-bar span within 3.2 percent of the stated excerpt
Drum material to repair fromYours s6685613 The densest drum candidate with the closest calculated duration. Density and duration and file-rule compliance, and no evidence yet on accuracy, cymbal choice, orchestration or attack placement
Nothing yets5820647 Flattened to 4/4, labels no source shows, and two of its five staves hold nothing but rests
The next move on this song. Align the community bar map against the exact album master first, then move your drum material into that structure bar by bar. Re-barring before that alignment is confirmed would carry any error in the map straight into the result.
On the parenthesised notes. A parenthesis is a normal ghost-note convention in published drum notation. Counting them records a house file rule and does not by itself mark an error. Measured against the flam test (a short preparatory stroke on snare or tom with a main strike following), 357 of the 359 parenthesised drum notes across these tabs are true ghosts sitting on their own subdivision. Only 2, both in Montana, are genuine flam candidates. Converting the rest to grace notes would misread quiet sixteenths as flams.

Pass 2: Penguin In Bondage

Chosen for the deepest published coverage on either record: eleven notation images across four transcriptions, one of them by a member of the band.

Published sourceCoversTranscriberSignature labels
Guitar-solo openingopening barsMike Keneally4/4  (♩= ca 70, Dm)
Album version2:09 to 2:33Kasper Sloots4/4, 3/4
12-9-73 early show1:42 to 2:49Kasper Sloots4/4, 7/8
12-10-73 late show2:39 to 2:59Kasper Sloots4/4
Combined published labels: 4/4, 3/4, 7/8. Both ambiguous signatures were resolved by zooming the image rather than guessed: the album one is 3/4, and the early-show one is 7/8.
Songsterr tabBarsBPMPartsPublished labels presentLabels written
s5469710
AI run
251115 / 1555 2 of 34/4 ×214, 2/4 ×24, 3/4 ×13

Per-part notes, rests excluded: Lead Guitar 619, Rhythm Guitar 1,992, Bass 769, Drums 1,547, Vocals 172.

The honest limit on this one. The 7/8 appears only in the 12-9-73 early show transcription, and the Songsterr tab is of the album version. Which Roxy night the album take comes from is not established here, so the missing 7/8 may belong to a performance the tab never claimed to cover. Scoring it as a flat miss would overstate the evidence. What does stand without that caveat: the tab writes 2/4 twenty-four times, and no published source shows a 2/4 anywhere.
A tempo question left open. Keneally writes the solo at ♩= ca 70; the tab runs 115 and 155. 155 is close to 70 doubled, so the tab may be notating double-time, the same choice seen on Zomby Woof. Settling it needs the audio.

Pass 3: Village Of The Sun

This one changes what the comparison can even claim. The published set covers three different performances, and each has its own meter map.

Published sourceTranscribersSignature labels
Roxy & Elsewhere, 1974 (the album), end section Paul Strawser, Wolfgang Ludwig, Kasper Sloots4/4
Helsinki 1973, openingSloots, Ludwig 9/8, 4/4, 6/8, 5/4, 11/4, 3/4
Helsinki 1974, openingSloots, Ludwig 12/8, 4/4  (metric modulation ♩=♩.)
Each performance has its own meter map. Helsinki 1973 carries 9/8, 6/8, 11/4 and 5/4. The album take is written in 4/4. Any score of "does this tab match the published labels" is therefore conditional on the tab and the transcription covering the same night. That generalises the caveat raised in pass 2 rather than being special to this song.
Songsterr tabBarsNotatedDrum notesLabels writtenNote
s412175
editor-made
432:06 4024/4 ×39, 6/4 ×4 half the record: 2:06 against 4:16
s6109180
AI, "LIVE"
2394:32 1894/4 ×238, 2/4 0 notes on Lead Guitar and Vocals
s6628550
AI, "live and better"
2614:32 1,7994/4 ×254, 2/4 ×4, 6/4, 5/4, 3/8 fullest: 2,612 rhythm, 991 bass, 492 vocal

Album length 4:16. Both AI tabs notate 4:32, about 6 percent long. Note counts exclude rests.

What holds without a version caveat. The editor-made tab covers roughly half the record, so a full-song score is the wrong yardstick for it. s6109180 is the third tab in this survey shipping two staves at 0 notes. Only s6628550 carries a full band across the whole length.

Pass 4: Cheepnis

Version check first, as pass 3 requires. The published transcription is labelled "Cheepnis, opening" and links back to Roxy & Elsewhere, so it is the album take. The tab notates 6:29 against the album's 6:34, 1.3 percent short, which is consistent with the same take. Its one synced video returned nothing when queried, so the version match comes from the duration and the page label rather than from the video.

Published sourceCoversTranscriberSignature labels
Roxy album, 7 imagesthe opening onlyKasper Sloots, Fall 2019 4/4, 12/8, 7/8, 7/4
Songsterr tabBarsNotatedPartsPublished labels presentLabels written
s5469732
AI run
2006:295 2 of 4 4/4 ×185, 3/4 ×7, 7/8 ×4, 6/4 ×2, 5/4, 2/4

Per-part notes, rests excluded: Lead Guitar 192, Rhythm Guitar 3,064, Bass 1,068, Drums 2,266, Vocals 779. All five staves carry notes.

The first AI tab in this series to carry an odd meter the published score also shows. It writes 7/8 four times, and the published opening uses 7/8. Zomby Woof and Penguin In Bondage both missed the odd meters entirely.
What it lacks. No 12/8 and no 7/4 appear anywhere across its 200 bars, and the published opening uses both. The tab's first meter change is at bar 59; the published opening is in 12/8 by its bar 5.
The coverage limit, stated rather than scored away. The published excerpt stops after the opening while the tab runs the full 6:29. The 3/4, 6/4, 5/4 and 2/4 the tab writes later have no published counterpart to check them against, so they are neither confirmed nor contradicted here.

What is ready for the next passes

Every song on both records that has published notation in the archive and at least one Songsterr tab.

SongAlbumNotation imagesSongsterr tabsComparable
Camarillo BrilloOver-Nite Sensation52yes
Dinah-Moe HummOver-Nite Sensation70no
Dirty LoveOver-Nite Sensation51yes
Fifty-FiftyOver-Nite Sensation62yes
I'm The SlimeOver-Nite Sensation35yes
MontanaOver-Nite Sensation83yes
Zomby WoofOver-Nite Sensation83yes
Penguin In BondageRoxy & Elsewhere111yes
Pygmy TwylyteRoxy & Elsewhere64yes
Village Of The SunRoxy & Elsewhere83yes
Echidna's ArfRoxy & Elsewhere45yes
Don't You Ever Wash That ThingRoxy & Elsewhere22yes
CheepnisRoxy & Elsewhere71yes
Son Of Orange CountyRoxy & Elsewhere54yes
More Trouble Every DayRoxy & Elsewhere32yes
Be-Bop TangoRoxy & Elsewhere51yes
All SkateRoxy & Elsewhere30no
Dickie's Such An AssholeRoxy & Elsewhere40no

Penguin In Bondage has the deepest published coverage at 11 images. Echidna's Arf and I'm The Slime have the most tabs to separate, at five each.

Method

Songsterr notation comes from the public meta API and its CloudFront path, with no browser and no login, which is the route recorded in the songsterr-tab-guide skill. Published notation comes from the Zappa drum archive built 28 August. Meters, tempo maps and note counts are read from the tab JSON directly. Nothing here is read off a rendered screenshot.

Two repairs were needed to the archive during this pass. Its Zomby Woof page had been saved as a Cloudflare interstitial rather than content, and 21 other pages had the same problem at the tail of the crawl, so 13 real pages were refetched. Separately, the first crawl kept images only from pages that mentioned drums, which left most per-song notation for these two records missing: 113 further images were fetched and validated.