Finalize local file deliveries with complete evidence

Choose the required deliverable

The edited artifact must answer the actual task. A player needs persisted Title values; a CUE sheet needs separate real audio tracks; a review copy needs finalized revision structures rather than a hidden-markup setting. Keep the originals and the complete reports.

Solve the actual player sorting task

The source user has files named <Track> - <Title> - <Artist>.mp3 and wants the stored Title to include the track prefix and .mp3 suffix. A filename rename alone cannot fix a player that reads ID3 Title. T277 writes TIT2, TPE1 and TRCK in downloaded MP3 copies, leaving the selected originals untouched.

Choose delimiters and legacy policy deliberately

Each nonfinal placeholder stops at the first following literal delimiter; the final placeholder consumes the remainder before the suffix. Every input token must occur exactly once, separated by nonempty literals. If a title itself contains the delimiter, choose another explicit filename convention; the tool does not infer the intended artist. Title output tokens may repeat. Track1–9999 is numeric for TRCK; its original zero-padded spelling is available to the Title template.

Strip removes an ID3v1 trailer. Sync updates its Title, artist and track using strict Latin1: both text fields must be≤30 characters and track1–255. Unicode α or 中 fails sync rather than silently changing or truncating it. A new v1 record uses genre255. Existing album/year/comment bytes remain, except the v1.1 track byte pair.

Verify real edited copies

The checked200-file batch yields200 MP3 copies and200 before/after rows. Independent Mutagen reads every persisted composed Title, artist and numeric track. Original MPEG bytes and their SHA256 remain identical; APIC artwork and TXXX custom fields retain their raw frames. A separate v2.4 Unicode example keeps two artwork and two custom-text frames without deduplicating their IDs.

• Download the ZIP and complete JSON/CSV; the table shows only the first200 rows.
• Only ordinary unflagged ID3v2.3/v2.4 is accepted. v3 output text is UTF16LE+BOM; v4 is UTF8.
• Duplicate edited frames, inner-NUL multi-values, v3 slash-separated artists, unknown flags, compressed/encrypted/unsynchronized frames, extended headers and footers fail. Other opaque frames may repeat and remain byte-identical.
• Limit200 files,30MiB each,80MiB total; tags1MiB each and complete outputs100MiB. Any failure or cancellation discards the batch output; correct the named input and rerun.

Split one recording from explicit CUE indexes

T278 accepts one matching FILE, increasing AUDIO tracks1–99, quoted TITLE/PERFORMER and INDEX00/01 at75fps. Album titles can precede FILE; printable ASCII REM is an opaque ignored comment. Other directives or multiple files fail. Time maps to the nearest sample with floor(cueFrames×sampleRate/75+0.5). Choose previous to append a next-track pregap to its predecessor, next to include each pregap at its own track start, or discard to remove those frames. Leading frames outside the first chosen start are separately counted.

The checked10s/8kHz recording has tracks0,2 and5+1/75s and second INDEX00 at1+50/75s. Previous gives ranges0–16000,16000–40107,40107–80000; discard removes2667 frames. All retained PCM16 WAV samples equal the source bytes. Other WAV depths and FLAC/MP3 use round(actualPCM×32768), clipped to[−32768,32767], with no gain normalization. Output is WAV, not source-format preservation.

The real rawFLAC sample has18 valid encoded frames and80000 PCM frames; native afconvert recovers the original WAV exactly. Both browser PCM snapshots were independently quantized and checked. FLAC requires known nonzero STREAMINFO sample count, unchanged rate/channels/depth and actual CRC8/CRC16/frame-structure validation; metadata is bounded8MiB/1000blocks and reserved metadata/unknown totals/Ogg FLAC fail. Its STREAMINFO MD5 is not separately authenticated. For the same48kHz MP3, Chromium supplies96000 frames while WebKit supplies97920 because of priming/gapless behavior; the final track follows the actual decoder end.

• Exactly2 selected files: audio30MiB and CUE1MiB;31MiB total,80MiB decoded float32 and100MiB complete output.
• Review every half-open frame range and dropped interval, not just the200-row table.
• A codec or CRC failure is a failure; no fallback output pretends to be a successful split.

Finalize supported DOCX structures instead of hiding review markup

T279 preflights the whole ordinary unsigned/nonmacro/nonencrypted DOCX before writing. It handles direct paragraph wrappers around text runs in main/header/footer/footnotes/endnotes. Accept unwraps insertions and removes deletions. Reject removes insertions and restores delText as normal text. Classic comments, actual target parts/relationships/content types and trackRevisions/trackMoves/trackFormatting settings are removed; untouched parts keep their original bytes.

The checked five-story file has5 insertions,5 deletions,3 classic comment markers and1 tracking setting. Independent python-docx/lxml reads new α/current for accept and old 中/previous for reject across the proper stories. Actual macOS QuickLook renders the current simple source-final view and accept with equal pixels; reject is visibly different. This simple local proof is not a universal Word/LibreOffice layout guarantee.

• Complex fields/property/move/paragraph-mark/table-cell/customXML/conflict/nested revisions, wrapper-local namespace bindings, revised nontext objects and modern threaded comments/people cause whole-package refusal—even inside a branch that the chosen policy would discard.
• One30MiB DOCX;80MiB expanded/10000 entries,total XML20MiB/1m elements,each XML200k elements/depth128,UTF8 XML1.0. Correct unsupported structures in a full editor and rerun.
• Metadata and orphan media may remain. Revision finalization is not anonymity or forensic erasure; retain original review copies separately.

Tools used in this article