BitShovel.One real problem at a time
Open the current workflowCurrent Workflow中文

One real problem at a time

中文

HISTORICAL DIG · JUL—AUG 2026

When is a cut ready for someone else to watch?

A downloaded shot proves only that a file exists. Dialogue can double, a character voice can drift, automatic captions can reverse a negation, and loudness or export faults can appear on another device.

First recorded · Last public revision · Revision 8
Picture lockAudio provenanceCaption reviewTwo-pass QC

AI SHORT-DRAMA MOTHER WORKFLOW · NODE 06

Relation to the current subject

Node 05 passes accepted shots; node 06 freezes one review master another person can watch.

Lock picture, then assign every line, ambience, effect, music cue, and caption one source and activation state. Two full playbacks answer different questions so a technical repair cannot silently rewrite the story.

Record state at the time

  • AI short drama
  • Audio, edit, and QC
  • Standard record
  • Developing
  • Historical E2 · maker method, technical standards, and specific issues rechecked · no BitShovel master

R8 CURRENT JUDGMENT

Judgment at the time

Playable proves a file exists; two clean QC passes make a review master ready for handoff.

Open full rationale and return rules
Node 06 locks picture before mix; it does not add generated shots during audio repair. Every spoken line gets exactly one active primary source: model-native, reference audio, post-recorded dialogue, or a caption-only fallback. Inactive tracks are explicitly muted rather than merely too quiet to notice. The voice sheet records character, source, version, permission, line, and time range; music and effects also resolve to provenance. Playback one checks story and identity: continuous action, one speaker, acceptable pronunciation and lip sync, and captions that preserve meaning. Playback two checks only the technical master: no doubled dialogue, clipping, or accidental mute; target loudness and true peak recorded for the intended handoff; a parseable, readable caption file; and complete playback in two independent players. Any failure keeps its timecode and returns to the named track or upstream node. Only two clean passes freeze the review master, captions, provenance sheet, QC record, and file digest for node 07. This node does not make channel upload packages or prove that the work is good, publishable, understood, or commercially viable.
Who it was for
People who leave node 05 with accepted shots and a cost receipt and must bind picture, dialogue, audio, captions, and file identity before a stranger watches
Why it mattered then
Open note

The danger in post is not that every fault is large; it is that every track looks almost ready. Without one active path, a speaker per line, openable provenance, and a frozen file digest, one fix can silently create another version.

Where the judgment stops
Open note

This is an unrun, unfrozen review-master method—not a BitShovel film, mix result, loudness certificate, caption-accuracy score, platform acceptance, or audience outcome. Every timecode, track, line, and state in the visual is fixed fiction and maps to no real person, work, voice, or file.

BEFORE OPENING THE POST TIMELINE

Fit check

If picture still moves or any critical sound lacks provenance, do not patch the gap inside the mix.

This record may help

  • Node 05 delivered accepted shots, exact versions, and a cost receipt, and picture order can now freeze.
  • Dialogue, music, effects, and captions can each name provenance, permission scope, owner, and an openable location.

Do not use it yet when

  • The script, shot, character asset, or generation route still needs to change; return to nodes 02–05.
  • Identity, consent, music permission, or use scope for critical audio is unknown; a temporary file cannot be mixed into a formal master.
  • The goal is platform upload, audience comprehension, or revenue validation—each belongs to a later node.

HISTORICAL TRIAL · UNRUN, UNFROZEN 120-MINUTE ACTIVE-POST CAP

Historical trial

Lock picture and primary dialogue first; then separate story-identity QC from technical-file QC.

Time needed
A 120-minute active-post cap including two full playbacks; render and waiting time stay separate; this round remains unrun
Likely cost
Use cleared tracks and existing edit tools first; list new voice, music, effect, plugin, and repair spend separately rather than hiding it in sunk subscriptions
Permission boundary
Handle only cleared shots and audio. Shared QC uses fictional IDs and digests, never real accounts, identities, contracts, keys, or unreleased dialogue
  1. 01

    First step

    Open this step

    Freeze picture. Assign every line one character, time range, primary audio source, and active state, then explicitly mute every alternate dialogue track. Add ambience, effects, music, and captions with provenance. Playback one checks story, identity, pronunciation, lip sync, and caption meaning. Playback two checks doubling, silence, clipping, target loudness and true peak, caption parsing, filename, digest, and complete playback in two players.

  2. 02

    What should exist

    Open this step

    One digested review master, one parseable caption file, one audio provenance and activation sheet, and a two-pass QC record with timecodes, owners, and decisions.

  3. 03

    How to tell it worked

    Open this step

    Both full playbacks complete. Every line has one primary source, all audio resolves to provenance and permission, captions preserve meaning, technical fields and digest are complete, and two players reach the end.

  4. 04

    Stop when

    Open this step

    Stop when a critical fault requires regeneration or rewriting, identity or pronunciation cannot be confirmed, a caption changes meaning, audio rights are unknown, or the next repair crosses the post cap.

  5. 05

    How to back out

    Open this step

    Keep the locked sequence, issue timecodes, and last playable export. Fall back to mute or captions for audio gaps, return shot faults upstream, and never patch a hole with uncleared material.

BITSHOVEL R8 MASTER RECEIPT · FIXED FICTIONAL TRACKS

Method and interfaces

One meaning-changing caption returns the caption track; a playable video does not earn a master digest.

Open method boundary

The visual contains no real editor interface, person, voice, or project. It shows how one primary dialogue source, explicit mute, two-pass QC, a timecoded issue, and an empty master digest fit together.

Playable is not ready for handoffBITSHOVEL R8 EXPLAINER · FIXED FICTIONAL QC RECEIPT
A fixed fictional review-master receipt with locked picture, post dialogue as the single active source, native dialogue explicitly muted, story and identity passed, and a meaning-changing second caption returning the cut instead of freezing a master

DEMO CUT 06 shows node structure only: picture is locked, every line has one primary dialogue source, and alternatives are explicitly muted. Pass one clears; pass two finds a caption that changes meaning, so the state is `return captions` and the master digest remains empty. It is not a BitShovel edit or QC result.

About this image and its source

Created by BitShovel on August 17, 2026 with no real person, voice, dialogue, work, file, software interface, account, or run result.

Open source page ↗

SOURCES AND EVIDENCE

Sources and evidence

A maker post method, loudness and caption standards, and two specific audio issues explain why provenance, activation, and QC stay separate. None establishes a BitShovel master or quality result.

Open evidence boundary

Historical E2 remains the old-page label. Revision 8 uses Runway's current multi-character method to show that per-character performance and local compositing remain post work, EBU R128 to show that measured loudness needs a declared target, and W3C guidance to show that automatic captions are only a starting point. ArcReel #1435 and #1486 remain two traceable overlap and voice-consistency issues, never a frequency estimate.

  1. 01
    Maker post-production method · Rechecked

    Runway's multi-character method still uses per-character performances and local compositing

    Open source boundary

    The current guide requires separate character performances, a base video, crop and overlay, and local editing. It shows post work remains after generation, not that any output meets BitShovel's bar.

    Open original source ↗
  2. 02
    Official technical recommendation · Rechecked

    EBU R128 separates programme loudness, loudness range, and true peak

    Open source boundary

    It supports recording a measurement target and descriptors in QC. It does not make one value universal across web platforms and is not a BitShovel measurement.

    Open original source ↗
  3. 03
    Official caption guidance · Rechecked

    W3C warns that automatic captions need human accuracy review

    Open source boundary

    The guide says automatic captions can start the work, while errors can change meaning. It supports line review, not a BitShovel accuracy score.

    Open original source ↗
  4. 04
    Specific user report · Rechecked

    ArcReel #1435: one export kept native audio and TTS together

    Open source boundary

    The issue supports explicit activation for mutually exclusive dialogue tracks. It establishes no cause, affected version, frequency, or BitShovel reproduction.

    Open original source ↗
  5. 05
    Specific consistency issue · Rechecked

    ArcReel #1486: one specification discussion asks for cross-clip voice provenance

    Open source boundary

    The discussion supports one sheet for voice source, version, and covered clips. It does not establish general voice drift in any product.

    Open original source ↗

REVISION RECORD

Revision record

R8 separates the review master from channel release packages and pins every failure to a track, timecode, and owner.

Open 8 revisions
  1. Revision 1

    Moved from animated clips toward a playable master with dialogue, sound, captions, and two-pass QC.

  2. Revision 2

    Added alternate tracks, voice provenance and line coverage, and traceable public audio issues.

  3. Revision 3

    The old page added speaker turns and per-channel exports but still mixed the review master with release packages.

  4. Revision 4

    Replaced the generated lead with a released-film frame carrying explicit authorship and an open license.

  5. Revision 5

    Moved the lead to a real source tied directly to audio post instead of reusing narrative frames for every production task.

  6. Revision 6

    Briefly restored an open-licensed narrative frame as the lead and moved post-production material downstream, still separate from BitShovel results.

  7. Revision 7

    Removed the unrelated narrative lead and used a traceable post-production desk, still only context rather than master or QC evidence.

  8. Revision 8

    Narrowed the node to picture lock, one primary dialogue source, provenance, and two-pass QC; left channel packages to the release node and replaced atmospheric photos with a fixed fictional return receipt.

WHAT THIS RECORD LEAVES

What this record leaves

Node 06 does not produce a video that merely looks done. It produces a master receipt the next person can verify.

Only two clean full playbacks, openable provenance and permission, meaning-preserving captions, and a frozen digest send the review master into node 07. Every failure returns to a named track or upstream node; no platform upload happens here.

Back to the Dig index