aesthetic.computer / core
Monorepo for Aesthetic.Computer aesthetic.computer
Star
Fork 0
Watch
core papers whistlegraph-platter
1 folder · 4 files
tools
dropbox-files.json
dropbox-manifest.json
dropbox-manifest.md
README.md
README.md

Whistlegraph Platter #

An index of whistlegraph materials across the repo, the web, and dropbox. A sub-platter within the papers platter — source material that feeds the arxiv paper, the are.na channel, the Annual Vol. 8 pitch, and anything else that comes next.

Whistlegraph — a drawing form invented in 2019 where every line is a sung syllable. You draw it while you sing it. The finished drawing is a score. The score teaches you how to play it.

This is index-only: pointers to where things live. No file migration, no binary duplication. Extend sections as material is cataloged.

1. In this repo #

Paper #

Practice surface (aesthetic.computer) #

Synced assets — per-piece score↔video pairs #

system/public/assets/whistlegraph/<piece>/ (1.0 GB, 47 files, synced 2026-05-02 from s3://assets-aesthetic-computer/whistlegraph/; assets dated 2023-10-21).

10 pieces with consistent structure — each has <piece>-score.png + <piece>-tt-compilation.mp4 + <piece>-web.mp4 + <piece>.webp:

  • butterfly-cosplayer · i-dont-need-an-iphone · lately-when-i-fly · loner · mommy-wow · people-pleaser · puzzle · slinky-dog · time-to-grow · whats-inside-your-heart

1 outlier piece — music-2-whistlegraph-2/: intro.mp4, poster.webp, spines.webp, .webp. No score PNG, no tt-compilation. Different shape; may be a music-track piece rather than a score-based whistlegraph.

Per-piece file roles:

  • <piece>-score.png (~400 KB – 1.8 MB) — clean graphic score, paper figure source
  • <piece>-tt-compilation.mp4 (~28 – 96 MB) — TikTok compilation: multiple takes of the same piece in one file. Primary signal for variation analysis (tempo / stroke-order / vocal phrasing across recordings of the same score).
  • <piece>-web.mp4 (~18 – 46 MB) — single web-friendly recording
  • <piece>.webp — poster for card-deck UI

Card-deck UI assets (top level): next-arrow.svg, play-circle.svg, play-triangle.svg

The 10 piece dirs ARE the score↔video pairs already organized by directory name. No further pairing work needed for these — the analyzer can iterate <piece>/ and have score + N-take video alongside. The tt-compilation files are the unfair advantage: each one is a built-in dataset of how the same score gets re-performed.

Standalone site #

Are.na Annual Vol. 8 submission (2026-04-20) #

2. External #

Whistlegraph-owned domains and accounts #

Institutional / commissions #

Press, prior writing, context #

Frequent collaborators beyond the trio #

Per Schneider Museum page: Ella Fleck, Niki Stebbins, Matt Doyle, Anastasia Lewis, Adam Schwarz, Ash Nerve, plus an online community of artists and fans.

3. Dropbox archive — /Whistlegraph/ #

Total: 12,903 files · 421.66 GB · 49 top-level subfolders (manifest pulled 2026-05-03 via Dropbox HTTP API).

Not locally synced — accessed via the API using the OAuth token in aesthetic-computer-vault/dropbox/auth.json (dbxcli format). Manifest tooling: tools/dropbox-manifest.mjs → outputs dropbox-manifest.md and dropbox-manifest.json.

Performance & recording #

  • Whistlegraph/TikTok/ — 233 files, 10.94 GB · 73 subdirs incl. Collabs/bo en/ (Bo En collaborations, late Jan–Feb 2021 .mov files)
  • Whistlegraph/YouTube/ — 278 files, 43.48 GB · gigabyte-class long-form (Improv.MOV 1.32 GB, relaxing-horror-nightmare-show-s1e1.mov 1.52 GB, IMG_7329.mov 3.38 GB)
  • Whistlegraph/Instagram/ — 14 files, 27.2 MB (thin — IG archive is small)
  • Whistlegraph/Live/ — 151 files, 11.45 GB · streaming assets (overlays, scenes JSON), live performance captures
  • Whistlegraph/Music/ — 1569 files, 23.84 GB · incl. Music/Goodiepal Recording Session/prut mixene/ (whistle montage WAVs ~200 MB each)
  • Whistlegraph/Audio Drafts/ — 1 file, 375 KB (camille run 22.7.12.m4a)
  • Whistlegraph/Recorder/ — 3 files, 519 KB (just UI assets — NOT actual recordings)
  • Whistlegraph/Misc. Videos/ — empty in the manifest pull (folder visible, contents not enumerated; investigate)

Scores & drawing source (primary analysis material) #

  • Whistlegraph/Scores/ — 39 files, 51.4 MB · Score Rev1 PDF subdirectory contains i-dont-need-an-iphone.pdf, its-time-to-grow.pdf, lately-when-i-fly.pdf, loner.pdf — same piece names as the paper figures and the per-piece DO Spaces dirs. Rosetta confirmed: each piece has a DO Spaces tt-compilation MP4 + a Dropbox Score PDF.
  • Whistlegraph/iPad Procreate Backups/ — 33 files, 6.38 GB · .procreate bundles with native per-stroke playback timing. Highest-leverage analysis input.
  • Whistlegraph/Pixaki Updates/ — 87 files, 280 KB · .pixaki projects (frame-by-frame; tiny).
  • Whistlegraph/Manuscripts/ — 22 files, 86.3 MB · per-piece manuscript pages (incl. slinky dog manuscripts/)
  • Whistlegraph/Research/Whistlegraphs for GAN/ — 78 files, 217 MB · already-curated JPEG dataset prepped for ML training. Pre-existing prior work — re-use.
  • Whistlegraph/Large iPad Procreate Update/ — 5 files, 824 MB · large procreate exports (Spiral_Smile.procreate 115 MB).

Exhibitions & press #

  • Whistlegraph/Exhibitions/ — 4426 files, 253.93 GB (largest single subfolder) · Rhizome Longest Whistlegraph Ever microsite media, manuscript selects
  • Whistlegraph/The Longest Whistlegraph Ever (so far) Media Archive.zip — single 19.24 GB zip at root
  • Whistlegraph/Whistlegraph Full Cover Ad Shoot/ — 218 files, 11.64 GB
  • Whistlegraph/Sex Magazine/ — 2522 files, 10.97 GB · zine production files (FINAL INDD, fonts, layout links)
  • Whistlegraph/Sex Magazine Launch/ — 101 files, 226 MB
  • Whistlegraph/NFTs/ — 14 files, 205 MB
  • Whistlegraph/Press/, Whistlegraph/Commercials/, Whistlegraph/Lectures/ (3.59 GB), Whistlegraph/Scholarly Articles:Academia/
  • Whistlegraph/Write a Whistlegraph Workshop/ — 43 files, 868 MB · social-media promo .mov files (Triangle, Factory, Puzzle)
  • Whistlegraph/Stamps/, Whistlegraph/Shop/

Community #

  • Whistlegraph/Media from Fans/ — 43 files, 635 MB · incl. The Wonderful World Of Whistlegraphs I: A Documentary.mp4 (166 MB)
  • Whistlegraph/Media from Friends/ — 301 files, 13.05 GB · large per-friend subdirs (e.g. From Katrín Helga Ólafsdóttir Whistlegraph Copenhagen Sept 3 2021)
  • Whistlegraph/Comments Screenshots/ — 1 file, 231 KB

Trio internal #

  • Whistlegraph/Whistlegraph Minutes/ — 4 files, 470 MB · zoom recordings (March 25 trio meeting w/ Ella, Camille, Alex, Jeffrey)
  • Whistlegraph/Hand Health/ — 8 files, 5.86 GB · large .m4v therapy videos
  • Whistlegraph/Moon Paintings/ — 16 files, 2.5 GB · collaborative paintings w/ Alex F. + CK (massive PSDs)
  • Whistlegraph/No Paint/ — 2202 files, 140 MB · the No Paint app source/assets (jeffrey + alex's earlier project)

Business (preserve, exclude from analyzer corpus) #

  • Whistlegraph/LLC/ — 6 files, 1.3 MB · Articles, EIN, Operating Agreement, W9, f8832
  • Whistlegraph/Invoices, Bills & Contracts/ — 19 files, 17.7 MB
  • Whistlegraph/Signed Documents/, Whistlegraph/Taxes/

Loose root-level files #

Bumbling Bumperfly.mp4 (29.9 MB), whistlegraph-tiktok.mov (7.1 MB), playbuttons.png (14.2 MB), playbuttonassets.ai, Whistlegraph Account Screenshot (Dark).jpg, Capture.PNG, Whistlegraph.png, Rhizome - Workshops Blog Post.zip (8.2 MB).

File-type signal (analyzer-relevant) #

Type Count Notes
.mp4 + .mov + .webm + .m4v 1551 performance video corpus
.wav + .aif + .m4a + .mp3 + .aac 1331 audio takes/stems
.png + .jpg + .jpeg + .heic + .webp + .gif 6225 stills, score scans, manuscripts
.procreate 40 per-stroke timing native
.pixaki projects (in Pixaki Updates) frame-by-frame source
.als (Ableton) + .aep (After Effects) + .prproj (Premiere) + .psd + .indd 268 composition/edit projects
.pdf 120 scores, manuscripts, contracts
.srt 26 subtitles (alignment ground truth)

4. Collaborators (2019–2023 trio) #

  • Jeffrey Alan Scudder — creator, aesthetic.computer, @jeffrey

  • Alex Freundlich

  • Camille Klein

  • Form invented by Jeffrey in 2019. Trio formed in 2020 (pandemic, cabin in Ashland, Oregon). Separated November 2023. December 2023, Jeffrey took solo access to all of the group's social media. The form continues under aesthetic.computer.

5. Timeline (rough spine) #

  • 2019 — Jeffrey invents the form. Early TikTok posts.
  • 2020 — Trio forms (Jeffrey + Alex Freundlich + Camille Klein) during the pandemic in a cabin in Ashland, Oregon.
  • 2020–2023 — TikTok growth to ~2.7M followers. No paid promotion. Viral hits with 4-stroke / 4-sung-line compositions.
  • 2021 — Rhizome commissions The Longest Whistlegraph Ever (so far).
  • 2022-05 — Ten Whistlegraphs at Feral File; The Longest Whistlegraph Ever (so far) debuts at New Museum as part of Rhizome's First Look.
  • 2023-06-06 — The Whistlegraph Zine published (Sex Magazine / Asher Penn), 60pp, 750 copies. Ciocci essay.
  • 2023-11 — Trio separates. Jeffrey continues solo.
  • 2024–2025 — Form continues under aesthetic.computer: /whistlegraph piece, self-documenting-score as AC's founding principle.
  • 2026-04-20 — Are.na Annual Vol. 8 pitch submitted (essay: Whistlegraph and the Self-Teaching Score; book Dec 2026).

6. Local off-repo caches #

Material on @jeffrey's laptop outside the repo, Dropbox, and Spaces. Useful for analysis without round-trips.

iCloud Drive (~/Library/Mobile Documents/com~apple~CloudDocs/) #

  • Downloads/whistlegraph-2023.1.30.02.30.20.mp4 — performance capture (Jan 30 2023)
  • Downloads/whistlegraph-2023.1.30.02.40.41.mp4 — performance capture (Jan 30 2023)
  • Downloads/whistlegraph-2023.2.04.15.35.23.mp4 — performance capture (Feb 4 2023)
  • Downloads/whistlegraph-2023.2.12.13.55.10.mp4 — performance capture (Feb 12 2023)
  • Audio Files/Noise_Shoegaze Whistlegraph.aif — composition draft
  • Audio Files/Noise_Shoegaze Whistlegraph v3.aif — composition draft (rev3)
  • Audio Files/Whistlegraph Practice.aif — practice take
  • Adam's Beats/4 2 whistlegraph 80 g .wav — collaboration with Adam Schwarz (80 BPM, key g)
  • 2021-03-21-Whistlegraph Retainer Agreement.pdf — retainer (LLC adjacent)

7. Analysis targets — what to pull, align, and structure #

Goal: feed a structural understanding of the whistlegraph form (stroke ↔ syllable ↔ time mapping). What we have so far is unaligned. To get to a learnable grammar, the corpus needs score-file ↔ video pairs.

Highest-leverage pulls (sized against the manifest) #

Sized "analysis core" subset is ~17.7 GB — Dropbox-only:

Priority Path Size Why
1 Whistlegraph/Scores/ 51 MB Rev1 PDFs match per-piece names from the paper figures + DO Spaces dirs. Clean vector source.
2 Whistlegraph/iPad Procreate Backups/ 6.4 GB 33 .procreate bundles with per-stroke timing replay. Skips audio-to-stroke alignment entirely.
3 Whistlegraph/Research/Whistlegraphs for GAN/ 217 MB Already curated by jeffrey for ML. Re-use, don't re-curate.
4 Whistlegraph/TikTok/ 10.94 GB 233 files, primary performance corpus. Score-as-video. Includes Collabs/bo en/ Bo En collaboration takes.
5 Whistlegraph/Manuscripts/ 86 MB Per-piece manuscript scans (slinky dog visible).
6 Whistlegraph/Pixaki Updates/ 280 KB Cheap; frame-by-frame source.
7 Whistlegraph/Audio Drafts/ 375 KB Single Camille take.

Already in hand (DO Spaces, locally synced 2026-05-02): 10 piece dirs with score PNG + tt-compilation MP4 + web MP4 + webp = 1.0 GB. No further pairing work needed for those 10.

External pulls (web, not Dropbox) #

  • TikTok profile crawl (yt-dlp against @whistlegraph URL or CDP against the creator dashboard) — supplements the Dropbox TikTok/ archive with anything missing or post-archive. The Dropbox copy is likely the canonical source given the upload patterns.
  • YouTube channel pull (yt-dlp against channel URL) — Dropbox YouTube/ has 43 GB of source files; the public channel has the encoded uploads. Pull whichever is easier per clip.
  • Are.na self-teaching-scores — 68 annotated blocks · already public · scrape via Are.na API for descriptions.

Heavy / situational pulls (defer until analyzer is working) #

  • Whistlegraph/YouTube/ (43.48 GB) — long-form, lower density per byte
  • Whistlegraph/Music/ (23.84 GB) — Goodiepal sessions, collaborative composition context
  • Whistlegraph/Live/ (11.45 GB) — streaming overlay assets, less score-density
  • Whistlegraph/Exhibitions/ (253.93 GB) — Rhizome microsite media; institutional context, not score corpus
  • Whistlegraph/Media from Fans/ (635 MB) + Media from Friends/ (13 GB) — UGC; analyzable as "responses to" the form, separate corpus

Pairing strategy (the "hedge") #

  • Score ↔ video pairing is what makes the corpus analyzable. Each named piece (e.g. butterfly-cosplayer, loner, puzzle) has a paper figure, a per-piece DO Spaces dir, and a TikTok upload. Cross-reference filename / title / hashtag to build the join key.
  • Procreate replays are the unfair advantage when present — per-stroke timestamps without alignment work.
  • Audio-to-stroke alignment for un-replayed pieces: vocal-onset detection (librosa/whisper) + frame-diff on the video; align local maxima within a beat-grid tolerance.

Are.na as ground truth #

The self-teaching-scores channel has 68 annotated blocks — human commentary per piece. Block descriptions are a free training signal for the "describe" step.

Out of scope for the analysis corpus #

  • LLC / Taxes / Invoices / Signed Documents / Hand Health folders — preserve in place, exclude from analyzer.

How to extend this index #

  1. When new material appears (a recording gets indexed, a press mention shows up, a new piece ships) — add a line here with a link or path.
  2. Use TBD: prefix for known-missing items so they're greppable.
  3. Don't check in large binaries; keep this as pointers.
  4. If the dropbox catalog grows large, split it into a sibling dropbox-manifest.md (still text, still pointers).