A folder named final, a duplicate marked 2, and an MP3 with no artist field can reveal more than a polished discography. Around Japanese experimental music—small-run CDs, cassettes, personal computers, and informal exchanges—surviving digital files often retain traces of circulation: who copied them, what software exported them, and which version was still unfinished when it left a hard drive.
None of this makes a file reliable by default. A downloader may have assigned the filename; a date may reflect a later copy rather than the recording session; a lossy encode may have replaced an earlier WAV. Still, this digital residue belongs in the record for noise, drone, ambient, electroacoustic work, field recording, and related practices that travelled through small, unstable channels. Treat it as evidence rather than trivia, while keeping its uncertainties in view.
What makes a digital file an artifact?
A digital sound artifact is more than its audio. It includes the traces around the file: its filename, directory position, format, duration, tags, checksums, accompanying text files, cover scans, transfer history, and the device or platform through which it was encountered. A cassette dub carries a physical history in its shell, leader tape, handwriting, and generation loss. A file has a comparable, though less visible, history in its metadata and technical structure.
This matters with experimental recordings because the release object may be incomplete, unavailable, or never have existed as a conventional edition. An artist might hand out a CDR at a performance, upload a short-lived archive, send a private download to collaborators, or circulate several altered exports of the same work. Such circumstances do not justify certainty about origin. They do call for careful description.
Separate the work from the copy
The first archival discipline is to distinguish between a musical work, a release, and the particular file in hand. They can overlap, but they are not identical. A 26-minute AIFF named room_take3 might be a master, a rehearsal document, an excerpt, or a renamed copy of something distributed elsewhere. Its sound may suggest a relationship to another recording, but it cannot establish authorship or release status on its own.
| Observed item | What it can support | What it cannot prove alone |
|---|---|---|
| Filename | A label used at one point in circulation | Official title or artist identity |
| Embedded metadata | Information present in that file version | That the tags were entered by the creator |
| Filesystem timestamp | A date associated with one storage event | Recording or original release date |
| Folder grouping | A collector’s or sender’s organization | Original release sequence |
| Audible match | Probable relation between recordings | Rights, authorship, or intended edition |
Read filenames as provisional evidence
Cryptic names are common in personal archives. Strings such as wasted_2ura, abbreviated catalog-like codes, or date fragments may point to a project, a side designation, or simply the habits of the person who copied the file. Roman letters and numerals do not necessarily indicate an English-language origin; they may reflect operating-system limits, software defaults, or a practical sharing convention.
Do not normalize an odd filename too quickly. In the preservation record, retain its original spelling, capitalization, spaces, punctuation, and extension. If a readable working title is needed, place it in a separate descriptive field. That way, a cataloguer’s guess does not quietly replace the source evidence. The blog’s examination of the orphaned Japanese noise cassette “wasted 2ura” is a useful reminder that a name can be suggestive without resolving a recording’s history.
Useful filename clues, used carefully
- Version markers: v2, final, mix, and master may refer to production stages, though people also use and reuse them casually.
- Side labels: A, B, ura (“reverse” or “back” in some contexts), and track numbers may suggest a cassette or CDR structure. Check durations and transitions before assuming a sequence.
- Date-like strings: Eight digits may resemble a date while actually being a serial number, transfer date, or unrelated identifier.
- Character encoding damage: Garbled Japanese text may result from an encoding mismatch. Preserve the damaged form and, where possible, document the decoding method rather than silently overwriting it.
- Extensions: A renamed extension can conceal the actual container. Confirm the format with a technical inspection tool instead of trusting the suffix.
Technical forensics without false precision
Basic inspection can reveal meaningful differences between copies. Format, sample rate, bit depth, codec, channel count, duration, tag fields, and encoder strings can help establish whether files probably descend from the same source. A 44.1 kHz, 16-bit WAV and a 320 kbps MP3 may contain the same performance, but they are different archival objects. The MP3’s encoder delay, changed high-frequency content, and tag history are part of its own copy history.
Checksums are especially useful. Generate a SHA-256 checksum for each preserved file and keep it in a plain-text inventory. A checksum does not identify a work, but it confirms that the exact bitstream has not changed during copying. If two files have different checksums, compare their technical properties and listen before deciding whether they are separate versions or merely different encodes.
Waveform and spectrogram views are diagnostic tools, not verdicts. They can reveal abrupt edits, long silences, broadband noise, clipping, spectral limits associated with lossy compression, or repeating loops. They cannot explain what a sound means, determine whether distortion was intentional, or identify its maker. In noise and electroacoustic work, a clipped waveform may be material rather than damage.
A practical intake procedure
When an unfamiliar archive arrives, resist the urge to retag, rename, or reorganize it. Those actions can erase the conditions in which the material was found. Make a stable copy first, then work from a derivative.
- Freeze the received structure. Copy the complete directory tree to preservation storage without changing names. Record the source medium, acquisition date, and any information supplied with it.
- Make a working duplicate. Use this copy for playback, transcoding, tag research, and folder experiments. Never let the derivative become your only version.
- Create an inventory. Include the full path, exact filename, file size, format, duration, checksum, and a short neutral note on audible content.
- Extract metadata separately. Save embedded tags and technical readings in a spreadsheet or plain-text record. This makes comparison possible without rewriting the file.
- Document interventions. If a file is renamed, repaired, decoded, or transcoded, record who did it, when, why, and which source file was used.
- Keep uncertainty explicit. Use phrases such as “attributed in filename,” “date reported by filesystem,” “audibly similar to,” or “unverified source.”
Retain the original audio file even if you create a more accessible listening version. Converting a low-bitrate MP3 to WAV does not recover missing detail; it only places the same audio in a larger uncompressed container. Label the derivative plainly: “WAV access copy made from MP3,” not “master WAV.”
Listening as comparison, not authentication theater
Close listening can connect files when database fields do not. Compare opening and closing sounds, the placement of silences, loop lengths, edits, channel behaviour, tape hiss, radio bleed, room resonances, and recurring processing gestures. In field recordings, a distinctive train announcement or rain pattern may suggest a shared recording environment without proving a location or date. In drone work, a file that appears static may reveal a slowly changing feedback path or layered source after level-matched comparison.
Level matching matters. A louder copy often seems fuller or more aggressive, particularly with harsh noise and compressed material. Compare files at similar perceived volume, and note observable features rather than aesthetic reactions: “contains 1.8 seconds of digital silence before onset,” “left channel drops at approximately 14 minutes,” or “same burst sequence appears with altered EQ.” Notes like these remain useful to later researchers without claiming more than listening can establish.
Context files are often more fragile than audio
Text documents, JPEG flyers, mail-order lists, scanned inserts, and old web-page exports may carry the only clues to a file’s circulation. Preserve them in their original form alongside the sound, including seemingly redundant duplicates. A README with broken line endings may retain a sender’s wording; an image filename may preserve a catalog code; a web export may show that a track was once presented as a stream rather than a downloadable release.
Private messages and email need particular care. They may clarify provenance, but can also contain personal information or reflect expectations of limited circulation. Archive the evidence needed for description without treating private correspondence as public liner notes. If a creator can be contacted, ask before publishing files, screenshots, email excerpts, or identifying details. Permission to possess a copy is not permission to distribute it.
Cataloging language that preserves doubt
The most useful catalog records separate observed facts from interpretation. Rather than writing “released in 2004,” write “folder timestamp: 2004; original release date unverified.” Rather than calling someone a “Japanese artist,” write “Japanese-language filename; creator not identified” when that is all the record supports. This is not timid cataloging; it makes the record auditable.
A compact record might include:
- Original filename and full path
- Exact file format and technical properties
- Checksum and preservation location
- Source medium or acquisition route, if known
- Embedded metadata transcribed verbatim
- Audible description using neutral terms
- Related files and the basis for the relationship
- Rights and access status
- Unresolved questions
That final field is important. A question such as “Does the suffix indicate a second transfer or a second piece?” gives later researchers a defined problem rather than a vague mystery. The investigation of cryptic filenames in Japanese noise archives, including “z3 1” shows why even a small label can gather competing explanations over time.
Access copies, preservation copies, and public claims
Collectors often face a tension between access and restraint. A fragile or obscure file may merit preservation, while a public upload could disregard an artist’s wishes or misrepresent unfinished work. Keep these decisions separate. A preservation copy may remain private; a listening copy may be limited to a research context; a public citation may describe an item without sharing the audio.
When posting an excerpt or discussing a file, state the level of evidence. Say whether information comes from embedded tags, a physical insert, an artist-confirmed source, a collector’s folder, or listening comparison. Avoid claims such as “lost release,” “official archive,” or “first digital edition” unless the documentation supports them. Ambiguity may be part of an experimental work’s history, but it should never become permission to invent certainty.
For a manageable first pass, take one folder and make a tab-separated inventory before altering anything. Record the folder name exactly, calculate checksums, save a screenshot of the original directory view, and keep one untouched copy. That small packet of evidence is more valuable than a cleaned-up library whose original structure can no longer be reconstructed.