NSFW Daily · 04
Change the world before you change the subtitles
Once two-person contact is inspectable, ask who the clip is for. Translate 闪婚, 气运, and 仙门 into English and keep the manhua face and the sect hall — English viewers are not watching the same film. Locale is a generation mode: visual grammar, place names, institutions, and the deal in the first five seconds all get rewritten. A translation layer will not do that.
Where this lesson sits
Daily 03 made two-person contact inspectable. This lesson is who it is for. Chinese, English, and Japanese are the same class: locale is a generation mode, not a translate switch.
The rule comes from a novel-to-video locale pack, not a slogan. Selecting `en-US` on the workbench changes visual grammar, narration voice, institutions, and whether BGM is on by default. It does not pipe a Chinese draft through a translator. The Daily keeps one shot you can shoot tonight.
Today's idea: the language pack rewrites the world, not the words
Chinese vertical manhua defaults to inked line, cel shading, cold grey for now, warm gold for flashback. An adult clip for US viewers is not that pack: harder anatomy, American-comic blacks, contemporary fashion and interiors — not hanfu, not a sect hall, not an idol-drama face. Translate the prompt and leave “manhua” in the style line, and you still generated a Chinese-audience world.
Institutions change too. Flash marriage is a webnovel beat; in English it needs a motive you can see — debt, board optics, a 90-day marriage of convenience, escrow. Do not keep 气运, 仙门, 丹房, or 道侣 as first-class places. Names are single tokens. Do not split a long name with spaces.
- Visuals: contemporary rooms and clothes for an English viewer; do not default to hanfu / sect / manhua face
- Institutions: escrow and a board, not fate-energy and a clan
- Time: one human clock; write rebirth as ages, not “lived sixty more years” plus “wakes at 22”
- First five seconds: the deal is visible in the frame, not an unread title card
One voice per shot; do not dice a voice-over into nine shots
Do not write narration and dialogue in the same shot. Lip-sync belongs to the person speaking; narration is VO, mouths closed. That is not a taste preference. It is the usual failure when mouth, voice, and picture fight.
Do not chop a continuous voice-over into timed mini-shots either. In a local audio run, slicing the same narration into a 9-shot prompt took word error from single digits to triple digits in the tail; one continuous line held. What you can ship is still the official 4–15s window. Longer audio is generated per semantic unit and joined, labeled as a concat — not a one-window minute.
Only the words changed
Translate the 闪婚 xianxia romance scene into English. Keep the hanfu hall, add English subtitles, cinematic, 4K.
Inspectable
Midtown suite, a 90-day contract between them, 6 seconds, no captions. An English viewer can see the deal without reading Chinese.
Two clearly 21+ synthetic adults in a Midtown Manhattan hotel suite at dusk. She sits on the desk edge in a black slip; he stands. Between them: an unsigned 90-day marriage contract, not a sect covenant. Neither speaks; there is no narrator. The deal must be readable from the frame — marriage of convenience, escrow, board optics. Locked camera, 6 seconds, breath and paper only. No captions, no hanfu, no xianxia hall, no third person, no flash-marriage imported as a mystic bond.
How to watch the file
First question: with sound off and no captions, is the deal readable in five seconds? Second: is the visual pack the one you declared? On failure, change the world. Do not add a longer English explanation.
- Are room and clothes contemporary urban, or did hanfu / a sect hall leak in?
- Can the paper, ring, or contract replace the words “flash marriage”?
- Captions, bubbles, site watermarks stealing the frame?
- Narration and dialogue in the same shot?
- Still Daily 03's two adults, or did a third person arrive?
- On failure, cut to 6 seconds, strip every letter, leave one prop that states the deal
Storyboard still means one inspectable shot at a time
Changing the world is not generating 24 episodes in one request. Each shot stays inspectable, same rule as Daily 01. Storyboard splits the story. Concat will not repair faces, and it will not rewrite a sect hall into Midtown.
Chinese viewers get vertical manhua grammar; English viewers get another visual and institutional pack. Do not mix both in one shot.
The age line does not move with the language
Characters stay synthetic adults, 21+. Changing language is not changing the age floor. Do not import a “senior disciple knocks her across the room” beat into an English adult clip as spice — US viewers read that as domestic violence, not a power fantasy.
This lesson is who you are shooting for. It is not how to evade review, and not how to “localize” an unauthorized real person.
Next lesson
Lesson 4 is who it is for. Lesson 5 will cover not splicing around a failed window: 5 seconds that cannot pass inspection must not be deleted to fake a master. Pixels under a mosaic are not recovered truth.
Until then, make the opening five-second deal readable from the frame.
FAQ
FAQ
If I translate the dialogue and add English captions, is that enough?
No. Captions are a translation layer. If visual grammar, institutions, and the opening deal stay Chinese, English viewers are still in a Chinese world. Change the room, the clothes, and the paper on the desk first.
Can flash marriage stay?
Keep the motive, not the unreadable institution. Make it a 90-day marriage of convenience, debt, board optics — a deal an English viewer can see — and put it on a prop in the frame.
Can narration and dialogue share a shot?
No. Narration is VO; only dialogue moves a mouth. Both at once makes lip-sync and voice fight. One voice per shot.
Why not cut the voice-over into many timecoded shots?
A sliced continuous VO loses intelligibility in the tail. Say one semantic unit in a 4–15s window; generate longer audio in segments and join them, labeled as a concat.
Does the same method apply to Japanese, Indonesian, Spanish?
The principle does: the language pack rewrites the world, not only the words. This lesson uses en-US because that visual and institutional pack is already written. Other languages still change room, deal, and the first five seconds before they translate lines.
Tonight
Write the five-second deal tonight
Open storyboard, schedule one shot, paste the prompt, generate 6 seconds. The deal must be readable from the frame — no subtitle explaining flash marriage.