Separate Instruction, Dialogue, and Subtitle Languages
| Layer | What determines it |
|---|---|
| Video instructions | The project’s chosen model dialect, prompt language, and creator direction; English is not mandatory. |
| Spoken dialogue | The script and sound direction. Label language, exact wording, and speaker per line. |
| Subtitles or explanatory translation | The delivery plan or reader’s needs. These do not replace spoken text. |
English camera instructions can accompany the script’s Mandarin dialogue. In an English tutorial, keep an explanatory translation outside the copyable verbatim line. Putting both versions in the generation instruction can turn one intended utterance into two. The skill explicitly separates instruction language from dialogue, voiceover, and off-screen speech, without guaranteeing compliance in every output.
Check Four Things per Line Instead of a Final “Chinese” Tag
| Check | Record |
|---|---|
| Speaker | A name or stable identity, rather than “someone speaks.” |
| Audio role | Visible dialogue, off-screen speech, or voiceover. Voiceover does not make the visible person speak. |
| Language | This line’s intended language or scripted dialect, not an inference from instruction language. |
| Verbatim text | Preserve negation, numbers, names, and conditions. Revise the script first if wording must change. |
This reader checklist is not the complete native video document. Add pauses, sighs, interruptions, or overlapping speech only when established; do not invent dialogue to fill sound layers. Between lines, identify the established ambience, silence, or the next speaker.
Original Comparison: English Instructions with Mandarin Speech
Fictional box-moving scene: Lin Yu stands left of the table and Zhou Yan right. Blue and red boxes are on the table, both people’s hands are free, and the established sound is air-conditioning with no music. Lin speaks in Mandarin; Zhou answers in Mandarin, then takes the red box while the blue remains. These are prompt plans, not generated-video results.
Incomplete: “English cinematic scene. Lin gives instructions, Zhou agrees. Chinese subtitles.” This omits both verbatim lines and the actual spoken language. A subtitle request does not supply the audio requirement.
Copyable English-language passage:
Locked two-person medium view, preserving Lin Yu on the table’s left, Zhou Yan on its right, and both boxes. Lin Yu, visible dialogue, spoken Mandarin, verbatim: “蓝箱留下，红箱搬出去。” Zhou listens without speaking. After Lin finishes, Zhou Yan, visible dialogue, spoken Mandarin, verbatim: “我只搬红箱。” After Zhou finishes, he lifts the red box with both hands and leaves; the blue box stays on the table. Between lines, retain the established air-conditioning ambience. Add no voiceover, other dialogue, or music. The complete speech, ordered movement, and ending must fit the actual shot duration.
For readers only, the lines mean “Leave the blue box; take the red box out” and “I’ll take only the red box.” Those explanations are not additional spoken dialogue. A Chinese instruction passage would retain the same two Mandarin lines, speakers, order, and sound roles rather than create translated utterances.
Check duration against full speech, pauses, carrying distance, and response. This page supplies no universal seconds or words-per-second rule. If the chosen production route does not support the required audio, choose another confirmed route instead of treating correct subtitles as correct sound.
After Generation: Locate the Error in Source Media or Editing
Listen line by line to the source video, then the edited output. Subtitles and transcription help locate speech; they neither override the script nor replace listening.
| Symptom | Next step |
|---|---|
| The source already uses the wrong language | Recheck per-line language and text, regenerate the affected material, and listen again. One added tag does not guarantee repair. |
| Language is right, but a negation, number, or condition changes | Mark the discrepancy against the script and repair that line or production plan. Wrong audio does not become a new script. |
| Source speech is complete, but the edit retains half a line | Check source speaking intervals and edit boundaries, preserving the beginning and end. Subtitles cannot supply missing sound. |
| Transcription is wrong while the audio is correct | Confirm by listening and correct subtitles or the transcript from the script rather than rebuilding correct speech. |
| Regeneration shifts subtitle timing | Measure the new source’s speech positions, update cut and subtitle timing, and inspect the output. |
Execute replacement audio, muting, or separate voice production with tools that actually support it, then inspect the output. The bundled editor preserves source audio and does not interpret CUT sound notes as automatic muting or replacement. “Use Mandarin dubbing” is not completed dubbing. Check lip sync after replacing visible dialogue; see the lip-sync guide.
FAQ
Do English prompts require English dialogue?
No. Instructions and spoken dialogue are separate requirements. Specify each line’s script language and wording, then inspect actual output. Reliability still depends on the production route and material.
Do it with the skill
Use $short-drama-video-prompts to inspect this passage while preserving the script. Separate instruction language from each line’s spoken language; check speaker, visible/off-screen/voiceover role, and verbatim text. Revise only the requested scope without translating dialogue automatically. List differences to listen for; identify actual tools for muting or audio replacement without reporting them as completed.
Read the method: Instruction versus spoken language · Per-line language, wording, and audio timeline · Exact dialogue, voiceover, and continuation · Source audio, transcription, subtitles, and mixing scope
About Drama SkillsThe short-drama-video-prompts skill on GitHub