Chat to R2V Prompt · Open Skill 公開スキル
Chat to R2V Prompt
Chat to R2V Prompt は Digen 上の公開 Open Skill です。開くと、このワークフローが何を作るか分かります。
Other
7 0 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4210253/2026-10-21/evebv62a.png
# Chat to R2V Prompt
The user talks like a human. This skill does the structuring. The user never writes a template,
never copies a field name, and never learns the format.
## When to use
- The user describes a scene, a cast, a location, an action, or a line of dialogue in plain language
and wants a structured prompt out of it.
- The user uploads reference pictures and wants them mapped to numbered subjects.
- The user wants an existing structured prompt revised, tightened, re-ordered, or re-formatted.
- The user types the keyword **"Convert to R2V Prompt"** (also accepted: "format R2V", "bikin
prompt R2V", "R2V prompt", "ubah ke R2V").
- The user asks "bantu bikin prompt" for a shot or a sequence.
- The user asks for a **dialogue and narration script document** for a film, a sequence, a scene, or
a shot ("bikin dialog", "tulis naskah", "dialog maker", "isi dialog film saya", "naskah narator").
- The user asks for the script as **audio**: plain TTS, TTS cloned from a voice source, or a direct
clone run ("buatkan versi audionya", "dub naskahnya", "pakai voice ini", "intonasi ceria/serius").
- The user wants the scene or the dialogue checked against **current pop culture, trends, or a
specific era and language register** ("cari yang lagi tren", "bikin nadanya kayak era 90-an").
This skill is **text only** by default. It never generates a video, and it never generates an image
unless the user explicitly asks to render a key frame afterwards. It also never generates audio
unless the user explicitly asks for an audio version of the script.
## Two modes
| Mode | When | Output |
| --- | --- | --- |
| **Video mode** (default) | the user describes movement, dialogue, or a sequence | the full field set, including shots, soundscape, and music |
| **Image mode** | the user says "buat gambar", "poster", "still", or describes one frozen moment | the same subject definitions, summary, retention analysis, and detailed description, but one paragraph, no shot list, no soundscape, no music |
If it is genuinely unclear, produce video mode and add one line offering a still version. Do not
ask an extra question just to pick the mode.
## The output skeleton
```
subject_definitions:
<Subject 1> is ...
<Subject 2> is ...
summary: [reference generation] The target video shows ...
retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - ...
detailed_description:
[style opening]
[Shot 1] ...
overall_soundscape: ...
non_diegetic_music: N
```
The six field names, their order, and the tag syntax are fixed. The wording inside a field is
free.
## Optional features
The features below are **add-ons, never the default**. The main function of this skill is turning
chat into a structured R2V prompt. The optional features are offered, not imposed.
| Feature | Trigger | Reference file |
| --- | --- | --- |
| **Dialogue maker** | the user explicitly asks for a dialogue and narration script, or accepts the one-line offer at the end of a delivery | `references/dialogue-maker.md` |
| **Narration script** | the user asks for a dedicated narrator or voice-over script, or a recorded voice-over | `references/narration-voiceover.md` |
| **Culture and era research** | the user asks for trend, pop-culture, or era grounding, or accepts the research offer | `references/culture-research.md` |
| **Audio version of the script** | the user asks for the script as audio, a dub, or a voice-over recording | `references/tts-audio.md` |
Rules that keep them from disturbing the main function:
- Never run the dialogue maker on its own initiative. After a normal delivery, end with **one
short offer line**, for example: "Kalau mau, saya bisa buatkan draft naskah dialog dan narator
untuk adegan ini juga." Deliver the offer as a single sentence, then stop. Do not start writing
the script.
- The script document is a **separate deliverable**. It never replaces, truncates, or rewrites the
six-field prompt, and it never changes the delivered promp