Reasons to Get a Bob · عوامی Open Skill
Reasons to Get a Bob
**Reasons to Get a Bob · Theme to Looks to Beat-Sync ref2v · Preset-Audio Native · Groove-Driven** *The agent skill for creators who want a finished **portrait transformation short**—not a product pack-shot reel, a mute mannequin lookbook, a talking-head explainer, or a six-clip stitch with drifting beats* **Reasons to Get a Bob** is a production-native Agent Skill for Digen. It orchestrates a complete **reference-to-video (ref2v)** pipeline for one viral-style vertical episode: **Idea → four look variants on one person → Groove Palette → one audio-driven ref2v cut → preset-track final mix** No storyboard grids. No contact sheets. No per-beat i2v chain. No dual-segment parallel batches. Music is **locked to the Skill preset** (`global_fixed_audio`), passed as `audio_url` at generation and mix. The body grooves to that track; hard cuts land on **2.63 / 5.50 / 8.33**. One episode, one video—HOOK plus three numbered looks inside a single 10s render. ## What you get | For creators | For the production stack | | --- | --- | | **Beginner-first** — no theme? the agent invents a look series and runs to the cut without stopping | **Opinionated pipeline** — four in-clip units, MCU lock, one identity, one room, preset audio, multi-ref prompt encoded in-skill | | **Transformation-first** — the plan is **what changes on the person and how they move on the beat**, not a silent mood board | **ref2v-first** — no accidental i2v / t2v / storyboard / per-shot video chains | | **Identity of face and room** — four person look plates plus optional user scene upload lock the host and background across the whole clip | **Audio-driven cut** — preset `audio_url` at `generate_video` and `mix_clips`; never put audio inside `input_urls` | | **Living groove grammar** — medium close-up, camera-facing, visible upper-body motion on every beat, same indoor frame in one coherent 10s timeline | **Dense prompts** — six-section English scaffold + time windows + hard-cut look changes + Groove Palette per unit | | **Delivery that continues** — after the cut, buttons are **Regenerate video / Change person or look / Stop**, already in your language | **Progressive disclosure** — craft lives in `references/`, not bloated context | | **Plain creator voice** — short lines about the theme, the looks, and the clip; no pipeline jargon unless you ask | **Marketplace-ready** — clear triggers, defaults, and tool contracts | ## Pipeline | Stage | Output | | --- | --- | | **Intake** | Lock language + **9:16** + **10s** + **single person** + theme labels (no orientation question) | | **Format DNA** | Locked MCU grammar, Groove Palette (four distinct motion families), four look plan | | **Variation plan** | HOOK → V1 → V2 → clipped V3 on Beat Map; PAYOFF beyond 10s dropped | | **Look plates** | Four vertical person stills (`t2i` baseline / `i2i` variants). User face or scene upload beats AI when provided | | **Beat-sync board** | One internal board — time / look / groove family / overlay / hard-cut hit | | **ref2v** | One six-section prompt + four look refs + preset `audio_url` → **10s** Beat-Sync Clip. Body motion follows the track | | **Mix** | Same preset track from t=0, video volume 0 → **Final Cut** with audible music | | **Regen** | Single-stage edits — new looks, new groove plan, or one new 10s video + mix | ## Production stack | Parameter | Default | Notes | | --- | --- | --- | | **Episode** | **10 seconds** | One `ref2v` video. Leftover song after 10s unused. User asks longer → explicit duration or more 10s chapters | | **Engine** | `ref2v` · `reference_to_video` · `rm3.5-ref` | Multi-ref look lock. Forbidden: `i2v` / `ai2v` / `t2v` / `fl2v` / dual-segment parallel | | **Clip duration** | **10 seconds** | Always set. Four units inside the clip are time windows, not separate videos | | **Video orientation** | Portrait · **9:16** | Landscape only if you ask. Look plates stay vertical even if the cut is landscape | | **Subject** | **Single person** | Theme = looks on that person (hair / clothes / pose / facing). Objects only as worn or held props in MCU | | **Image refs** | Four person looks required | User person and/or scene uploads preferred. No storyboard stills, no grids, no standalone room plates, no spatial composites | | **Ref order** | Person baseline, Look 1, Look 2, Look 3 | Subject labels in the prompt must match `input_urls` exactly | | **Music** | **Preset-locked, audio-driven** | `global_fixed_audio` as `audio_url` + `audio_start_time=0`. Forbidden: `generate_music`, user upload prompts, audio inside `input_urls` | | **Beat Map** | **2.63 / 5.50 / 8.33** | Hard cuts only. Never shift, compress, or invent hits. 11.10 PAYOFF not generated | | **Groove** | **Groove Palette · visible amplitude** | Four different motion families per run (shoulder pop, chest pump, figure-8, elbow pump, etc.). HOOK medium → V3 maximum. Forbidden: mannequin still, subtle-only sway, same sway template every run | | **Framing lock** | **MCU chest-up** | Locked-off camera, eye-level, person centered ~55–70% frame height. Forbidden: wide, product hero, new room per beat, footwork dance MV | | **Overlays** | **On** | Hook title `Reasons to get a {theme}:` plus numbered `1.` `2.` `3.` labels in your language | | **Dialogue** | **Off** | No TTS, no spoken lines, no lip-sync VO. Motion follows music, not speech | | **Cut grammar** | **Hard cuts inside one clip** | Look snap on the beat; groove never stops between units | | **Confirm** | **After delivery** | Default run auto-flows; show Beat-Sync Clip after generation, then Final Cut after mix | | **Language** | **User-typed lock** | Replies in your language; on-screen theme words match your language; image / video prompts are English | ## Modes | Mode | You bring | Agent delivers | | --- | --- | --- | | **Full auto-run** | Nothing — or “make a Reasons to Get a Bob video” | Invent a theme + four looks → groove plan → one beat-sync cut | | **Theme-first** | Topic + one-line premise | Expand to look series + overlays → full pipeline | | **Face-driven** | A person photo + brief | Bind the upload as identity; `i2i` look variants preserve the face | | **Scene-driven** | A room / interior photo + brief | Lock background to your scene; composite or generate MCU person inside it | | **Face + scene** | Person photo + scene photo | Use both via `i2i` — your person in your room across all four looks | | **Product as prop** | Product image + brief | Person remains the subject; product worn or held in MCU only | | **Single-stage edit** | “Change the looks” / “Redo the video” / “New groove” | That stage only, then one new 10s video + mix | | **Landscape on request** | “16:9 / horizontal” | Override the 9:16 default for this run only | ## Example prompts **Full auto-run** > Make a portrait Reasons to Get a Bob video, 10 seconds, rapid look changes on the beat. **Theme-first** > Theme: cottagecore haircut. Four looks, same person, same room, groove to the built-in track. **Face-driven** > This image is the person. Build a 9:16 transformation short. Theme: bold bob cuts. **Scene-driven** > Use this room photo as the background. One person, four outfit changes, beat-synced. **Face + scene** > This is the host, this is the room. Reasons to get a pixie cut — portrait, 10 seconds. **Product as prop** > The person wears this jacket across the looks. Still a person video, not a product ad. **Regenerate** > Keep the theme — redo the video with different groove moves. **Non-English overlays** > Reply to me in English, but the on-screen theme words stay in Spanish. **Landscape override** > Same format, landscape 16:9, 10 seconds. ## When to use - Vertical transformation shorts: haircut, outfit, style, identity, “reasons to get a ___” formats - One-line themes, uploaded face or room photos, or creators chasing the viral hard-cut look-change rhythm - Beat-synced **upper-body groove** on a fixed preset track — not speech, not a dance MV - Anyone tired of mute lookbooks, product tabletop heroes, Pixar ad pipelines, or four separate 10s clips stitched after the fact > **Not the right tool when** you need a talking-head presenter with TTS lip-sync, a 60-second six-clip mix, a product-only pack-shot with no on-camera person, custom user-uploaded music, generated BGM via suno, first-person POV walkthrough, storyboard grids for client sign-off, full-body footwork choreography, dual-host podcast cuts, or a drama series with many unrelated locations per minute. ## Invoke ``` /reasons-bob-shorts ``` Or: *"Turn this into a 10-second portrait transformation short — four looks on one person, preset beat track, hard cuts at 2.63 / 5.50 / 8.33, one ref2v cut."* **Your theme · Your person · Four looks · One grooving ten seconds.**
Custom
272 6 31
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3325120/2026-08-20/2m9ttnbg.jpg
---
name: reasons-bob-shorts
display_name: Reasons to Get a Bob
description: Create viral "Reasons to Get a Bob" transformation shorts with AI-powered hairstyles, outfits, and cinematic style changes.
allowed-tools:
- generate_image
- generate_video
- mix_clips
- recognize_image
- list_models
---
# Reasons to Get a Bob Production Team
Photoreal UGC short video. **The on-camera subject is always a person, never an object.** One identity. One room. Fixed MCU. Continuous look changes (hair / clothes / pose / body direction) on hard cuts. Beat-synced units inside **one 10-second** clip on the **Skill preset fixed track**. Incomplete leftover beats are dropped. No structured project state — progress is conversation + canvas assets + `write_todos`.
## Global Language Lock (Mandatory, Highest Priority)
[Language Protocol] Follow the language of the user's input; use that language throughout for replies, variation plans, storyboard tables, review conclusions, and guidance buttons.
### 0. Initial Language Lock (Init; triggered upon receiving the first message, must be completed before the first reply)
- On the user's **first message**, immediately lock `user_lang` and use it from the first reply onward. Greeting in one language then switching is forbidden.
- This Skill document is English (document language only). Branch English headings/templates are internal; translate user-facing text to `user_lang`.
- Before every user-facing output, self-check `user_lang`. If `user_lang` is English, include no Chinese characters.
- Do not switch language because branch files are written in English.
### Model Prompt Language (Mandatory; separate from user-facing language)
| Layer | Scope | Language |
|---|---|---|
| **USER-FACING LANGUAGE** | Replies, variation labels shown to the user, storyboard tables, `set_guidance` copy | `user_lang` |
| **MODEL PROMPT LANGUAGE** | Final **VIDEO GENERATION PROMPT** passed to `generate_video` (RM35 six-section structure in Branch 5) | **English always** |
| **IMAGE GENERATION PROMPT** | `generate_image` prompts in Branch 3 | English structural keywords + project-specific content |
| **ON-SCREEN TEXT** | Burned-in hook title and numbered labels | Match `user_lang` for the theme words; keep the format `Reasons to get a {theme}:` / `1. {label}` |
- **VIDEO GENERATION PROMPT must always be English** in all six RM35 sections. On-screen overlay wording inside `<d>…</d>` keeps the `user_lang` theme words.
- Global Language Lock **does not** authorize a Chinese video Prompt body when `user_lang` is Chinese.
### Internal vs User-Facing Output (Mandatory)
- **INTERNAL EXECUTION DATA** (never show): validation checklists, Asset ID Registry tables, World Logic Pass summaries, Beat-Sync board internals, routing/model parameters, environment/room stills, spatial composite images. Full list: Branch 2 Section 15.
- **USER-FACING OUTPUT**: short progress notes; **four person look stills** (optional during asset stage); the **generated Beat-Sync Clip** after video generation; the **final cut** after mix. Do not show room stills or spatial composites. Do not dump prompt packs or ask the user to confirm a traditional script.
> This document is the **master document**. When executing a given stage, **must first `read_skill_file(name='reasons-bob-shorts', path='references/…')`**. Do not guess branch content.
> **Before modifying this Skill, first `read_skill_file(name='reasons-bob-shorts', path='references/09-skill-editing-rules.md')`.**
> **When video generation is involved, first `read_skill_file(name='reasons-bob-shorts', path='references/10-forced-dual-segment-mode.md')` (Single Video Generation Mode).**
## What This Is
A production system for **one content type only**: Reasons-to-Get-a-Bob-style shorts.
**Format (this is the film, not a metaphor):**
fixed camera + person centered MCU + continuous costume/look variants + **beat-synced upper-body groove** + music hits + numbered captions.
Same person. Same indoor