Choose the sound’s job before adding it
A knock, key on a table or chime can mark completion, offscreen presence or a state change. Decide what viewers need to hear before selecting the sound instead of adding a generic impact to every action.
If the source has a clear knock, listen to that version first. Duplicating it can create a double hit or incorrect distance. When adding one, inspect audible onset, duration and space rather than trusting the filename.
Scene sound design addresses attention and story purpose; this page implements a chosen cue in the edit. Music, ambience and dialogue retain their own roles.
Separate source, cut-local and output time
| Check | Meaning |
|---|---|
| Source event | Where the action occurs in the source. |
| Cut in-point | Which source time was removed. |
| Local cue | Event time relative to the retained segment’s start. |
| Actual prior spans | Sum rendered segments rather than planned lengths alone. |
short-drama-edit mixes an 音效 cue at cut-local start/end times. Its implementation takes the specified duration from the sound file’s beginning before placing it in the film. The field is not an in/out range within the effect file. Prepare a source with suitable onset externally or revise the plan if it begins with silence; do not invent an extra offset field.
A tail needs room too. An over-short window can truncate reverberation despite correct onset. Choose or prepare an appropriate file, then update its real path and timing.
Complete example: placing a knock at source time 3.20
This original paper edit produces no sound file or actual mix. In a nine-second scene, a visitor knocks and someone inside answers, “Coming. Leave the stand by the door.” The preceding rendered cut lasts exactly 3.00 seconds in this example. This cut retains source 2.00–8.00 for 6.00 seconds. Contact at source 3.20 becomes local 1.20 and output 4.20; dialogue begins at local 2.00.
Assume no distinct original knock and a selected single-hit file sounding immediately at zero, with its required tail ending within 0.60 seconds. Write this cut-list excerpt:
## CUT-EP001-002 · Visitor knocks
- 来源：MOTION-EP001-002 · 制作成果/videos/SHOT-EP001-002.mp4
- 入点：2.00
- 出点：8.00
- 时长：6.00
- 取舍：保留敲门接触、完整对白与末尾反应
- 声音：保留原声与完整对白，补一次敲门；回听不盖字。
- 音效：1.20-1.80 制作成果/音效/单次敲门.wav（增益：-8）
- 字幕：无
Native field labels stay Chinese for the parser: source, in, out, duration, sound plan and sound-effect cue. Replace illustrative paths with real inspected media. The -8dB gain is an audition candidate, not a universal knock setting. The 声音 line describes intent; the tool parses 音效 for this operation. Intent alone does not add sound.
After mixing, follow the route: hand contact at output 4.20 matches the attack; no leading silence or second knock appears; the tail does not mask dialogue beginning at local 2.00; the whole sentence and final reaction remain. Removing another 0.40 seconds from this cut changes the source in-point to 2.40 and event-local time to 0.80. Do not retain the old 1.20 cue.
Frame boundaries can affect spans, so accumulate actual rendered durations and listen to the result. Correct paper arithmetic does not prove sound content, peaks or synchronisation.
Inspect the mix rather than the render exit code alone
The cited implementation mixes effects into the film’s original audio with optional effect gain. Adding inputs does not automatically divide the original audio level, and the 声音 plan does not execute dialogue ducking, muting or crossfades. Perform those externally when needed, keeping a project or processing record before checking output.
The tool’s short tail fade does not judge whether long decay survived; summing sounds can still become too loud or mask dialogue. Overall loudness processing follows effect mixing, with output measurement and listening still needed.
Brief the operator:
Add only this knock, preserving shot order, action outcomes, whole dialogue and other audio. Check current contact time, cut in/out and the file’s audible onset; update the effect cue and actually mix it. Listen through the sentence, joins and ending, recording output and remaining issues. If original audio suffices, add no duplicate. A sound plan is not a result.
For broader timing faults, see sync after editing.
FAQ
Why can a correct cue still sound late?
Check silence at the file start, whether the cue uses cut-local rather than source time, and actual rendered spans for output position.
Is adding the effect enough to deliver?
Check attack, tail, dialogue and summed peaks. Native mixing does not execute every sound-plan instruction; listen to and measure the final file.
Do it with the skill
Use $short-drama-edit for this cut’s specified effect with real media, cut list, chosen sound and trigger time. Update local cues, actually render and listen, preserving dialogue and recording any additional external mixing.
Read the method: Sound plans and native mixing boundaries · Cue parsing, placement and actual mixing · Cut-list example with native fields