Classify by physical location, not by a “phone sound”

Ask three questions for every line: Is the speaker visible now? Are they physically in this scene? Is this present-tense dialogue, or subjective/narrative speech from outside the scene?

Physical factCurrent native screenplay formRecording-handoff candidate
Speaker is visibleName: line, optionally with a performable cuesync, or the project’s accepted dubbed route
Speaker is in the same scene but outside frame[OS] Name: lineOS, or dubbed when production chooses it
Speaker is not present and supplies subjective/narrative speech[VO] Name: lineVO, or dubbed when production chooses it
Remote person joins the live conversation by telephoneEstablish the connected call and source, then use ordinary dialogue with an unambiguous speakerThis tutorial uses dubbed; other projects follow their actual recording and production choice

The last row is a safe handoff for this tutorial, not a new repository syntax. The pinned native format lists supported tags such as [VO], [OS], [SFX], [画面文字], [连续性] and [转场]; the recording checker accepts only sync, dubbed, VO and OS channels. Do not invent a [PHONE] tag, a phone channel, or a claim that telephone filtering has already happened.

The repository also documents another active production dialect that a project may adopt: Name (OS): is a deliberately chosen inner voice, while Name VO: is a deliberately chosen non-present or narrative voice; a telephone call, memory or explanation does not automatically become VO. Read which dialect the project has adopted and keep it consistent. Parenthesized OS/VO and native bracketed [OS]/[VO] are not synonymous labels and should not be mixed in one production script.

Let the call change a visible decision

Phone scenes often turn into alternating exposition. State what the on-site character intends to do before the call, give the remote person bounded information, and make the on-site character choose, verify or refuse. One remote line cannot establish that a repair is finished, an item has arrived or a relationship is restored.

Copyable five-field card:

On-site task: What must the visible character complete now?

Remote location and source: Where is the other person, and how is the call heard?

Bounded knowledge: What can the caller confirm, and what remains unknown?

On-site response: Who verifies, receives or objects?

Scene result: What has happened, and what still awaits execution?

Showing the caller may require an established video-call screen or a cut to another location; that changes shot and asset needs and cannot be replaced by a sound label. If the caller is heard only, the visible character’s pickup, pause, repeat-back or action should still show that the information was received.

Complete original example: moving costume rails during a leak

This fictional scene was designed for instruction. No audio, generated footage or edit has been tested. A west window leaks in a community theatre’s costume store. Coordinator Lin Qiao is on site, custodian Luo Min has gone home, and stagehand Tian stands just outside the open costume-store doorway. She is physically in the same scene, remains outside frame, and can hear the speakerphone on the table. Only a corner of the dry room is available before 18:20. A spare key may be sealed in an envelope at reception. The call must confirm the key record and which cabinet’s costume rails and contents move first, without pretending the window, full inventory or insurance work is complete.

INT. COMMUNITY THEATRE COSTUME STORE — NIGHT

Rain drips from the lower edge of the west window into a bucket. Lin photographs the wet area and writes cabinet A-03 in a temporary log. She stops between A-03 and A-04 without opening either.

Lin calls Luo, enables speakerphone and places the phone on a dry table.

LIN: The west window is dripping. A-03’s base is wet; A-04 is still dry. Which cabinet’s rails can I move first?

LUO (through phone): Move the covered rails and costumes inside A-03 first. The spare key is in a white envelope at reception marked “Costume Store Two.” Last night’s handover book records the seal on line seven. Do not move A-04’s contents; I have not seen their condition tonight.

LIN: I’ll search only for “Costume Store Two,” check the seal and line seven in front of a witness, then move A-03’s covered rails to the dry room. The window and A-04 remain for on-site review.

Tian remains outside frame just beyond the open store doorway. She hears the key instruction on speakerphone.

[OS] TIAN: The white envelope is in the reception box and the seal is intact. I’ll get it now and bring the box to the store door for your signature.

LIN: Good. I’ll sign for the key. Witness the cabinet number when I open A-03.

Tian leaves the doorway, then returns and enters frame with the sealed envelope and handover book on the dry table. They match “Costume Store Two” to line seven. Lin signs 18:05 and Tian signs the witness field. Lin opens the envelope, takes the key and unlocks A-03. Together they push the first covered rail toward the dry room.

LUO (through phone): After all A-03 rails reach the dry room, report the cabinet and item count. I’ll record your move notice, not fill in a count I cannot see.

LIN: Understood. We have confirmed the key handoff and started moving the first rail from A-03. The remaining rails and final count stay pending until everything is placed.

Luo is elsewhere. Her live phone dialogue is neither same-scene [OS] under the current native definition nor narrative [VO] merely because she is invisible. Tian stands just outside the open store doorway, physically in the same scene and able to hear the speakerphone while outside frame, so her first line is [OS]; after she enters frame, action and speech return to the on-camera form.

What has happened: the envelope and book reach the store, both records are checked, Lin signs for the key and opens A-03, and the first rail starts moving. The remaining A-03 rails and costumes have not all moved; the final item count, A-04 inspection, window repair and Luo’s receipt of the later report remain pending.

Project the same wording into the recording handoff

The recording sheet is not a second screenplay. Build or refresh the index from the current script, point each record’s source_ref to the real block, and keep line_text and speaker byte-faithful. Telephone spatialization, filtering and level belong to later sound implementation; a correct channel does not mean that work exists.

For the example, copy this handoff logic:

LineRequired handoff
Lin’s first linechannel=sync; addressed to Luo; knows A-03’s base is wet and A-04 appears dry; tactic=request bounded authority; visible lip sync required
Luo’s first linechannel=dubbed; addressed to Lin; knows only last night’s sealed-key record, not A-04’s condition tonight; tactic=limit the move; no on-camera lip-sync constraint
Tian’s off-screen lineScript block is [OS]; handoff channel=OS (or project-approved dubbed); just outside the open doorway, physically in the same scene and able to hear the speakerphone; tactic=provide a checkable on-site response
Luo’s final linechannel=dubbed; follows Lin’s request; tactic=limit what she can attest; item count remains unknown

Also preserve line_id, speaker, speaker_display, preceding_line_id, addressed_to, speaker_knows_now, tactic and any genuine target_seconds. After a script rewrite, regenerate the projection rather than polishing the sentence only in the recording sheet.

The pinned implementation checks the channel enum, conflicts between VO/OS blocks and channels, wording against the indexed block, and speaker disagreement. It does not judge convincing telephone sound, performance, distance, mix or generated audio. Listen to the actual recording and film after the structural pass.

Review source, listener reaction and result in the edit

Lock complete lines before choosing pictures. The phone need not remain visible for every remote line; the edit may show Lin checking the label, Tian entering and both signatures. Keep the telephone line complete and identifiable, and do not let Tian’s same-scene [OS] sound like another sentence from remote Luo.

Checklist:

  • A pickup action or established source precedes the first remote line.
  • Lin’s repeat-back does not add a claim that A-04 was inspected tonight.
  • Tian’s physical move from just outside the open doorway into frame remains continuous, with an identifiable speaker.
  • Receipt, unlocking and the first A-03 rail beginning to move are visible; the remaining contents and later item count stay pending.
  • Telephone timbre comes from the actual sound plan and output, not from a screenplay label.

If footage shows only Lin nodding, with no intelligible remote information and no repeat-back or action that carries it, new subtitle text cannot manufacture the missing call. Restore available sound evidence or request necessary production coverage.

Five phone-scene failures and repairs

FailureWhy it breaksRepair
Every phone voice becomes [VO]Live dialogue is mislabeled as narration and may receive the wrong recording constraintDecide whether the speech is truly subjective/narrative; otherwise establish the telephone context and use the adopted recording channel
A remote caller becomes [OS]Current native [OS] requires physical presence in the same sceneReserve [OS] for a same-scene off-screen person and state the caller’s remote location separately
Invented [PHONE] or channel=phoneThe pinned parser/checker implements neither valueUse supported syntax and sync/dubbed/VO/OS; leave telephone treatment to the sound plan
The sheet quietly rewrites a lineScript and recording lose one authority, splitting caption and performance versionsChange the screenplay, then rebuild the index/projection; record unresolved differences rather than overwriting
The call only recites backgroundThere is no visible receipt, leaving the edit with passive listeningMake the on-site person verify, refuse, sign or start one bounded action, and state what remains pending

FAQ

Is every telephone voice off-screen dialogue?

No. Current native [OS] means a person physically in the same scene but outside frame. A remote caller does not meet that condition and does not automatically become narrative [VO]. Establish location and source, then use supported dialogue and recording channels.

Can I add a `[PHONE]` tag for the tool?

Not in the pinned implementation. Put telephone state in scene facts, performable cues and the sound handoff. Parser extension would require a separate code change and tests; this article does not pretend it exists.

Do it with the skill

Use $short-drama-write for this phone-call scene. First state each speaker’s physical location, visibility, how the call is heard, current knowledge and the action the call must change. Preserve the project dialect and add no [PHONE] tag. Deliver the complete scene plus a verbatim recording handoff, separating structural checks from performance review.

Read the method: Current screenplay syntax and VO/OS definitions · Dialect boundary: a phone call is not automatically VO · Verbatim recording handoff and validation scope · Implemented channel-to-screenplay checks

About Drama SkillsThe short-drama-write skill on GitHub

All guides