Candy Series Director · Open Skill 公開技能
Candy Series Director
Director Mode - Script to Screen - 16:9 Native The agent skill for creators who want a finished 15-second cinematic clip—not a suite of disconnected frames. Movie Maker is a production-native agent skill for Veo2. It orchestrates a complete reference-to-video pipeline: script → world & shot plan → reference media → export ref-to-prompt → Veo render No storyboard grid. No shot-by-shot image frames. One cohesive short-form clip where characters, locations, and props stay locked across motion and style. What you get | Per-session | Per 15-second video asset | |---|---| | One-pass production—say "just make it" and the agent runs the full stack | Turnkey video—player-reference, locked continuity, production-ready | | Material identity-lock—ref images (text, artifact, character, world) stay persistent | ref-to-prompt—no storyboard grid—full sequence spec | | Tactile design—no "generic AI look"—heavy rhythm, continuity for short-form clips | Progressive depth-growth—controls lock in—ref assets expand, not blanket control | | Structured sequence setup—chapter concept and production-ready story | Showrunner-ready—shots progress, style holds, layout stays solid | Pipeline | Stage | Input | |---|---| | Input | Premise, characters, dialogue, story outline. | | Planning | World setting, visual style, shot list (1–2 total sequences). | | Reference Asset | Character / location / prop images (0–1 per asset type). | | ref-to-prompt | Merges references into coherent, unified 15s director blueprint. | | Render | One 15s video via 16:9 production render. | Production specs | Parameter | Default | Notes | |---|---|---| | Engine | veo1-p / reference_image_to_video | Single-step ref-to-video pipeline render | | Duration | 15 seconds | Built to fill the 15-second window | | Mode | 2–3 shot sequence | Authorial director clip with progression | | Orientation | Landscape | Landscape-first output | | References | up to 1 per category | Locked across image and video tasks | | Output | 15 second video + caption | Community-ready presentation output | Modes | Mode | You bring... | Agent delivers... | |---|---|---| | Full auto run | A concept or story idea | Script → spec → refs → render | | Script first | Your script or partial draft | Plan → refs → render | | Reference-driven | 1–3 ref images + brief | Spec uploads plan → render | | Single-shot edit | "Change dialogue" / "Adjust rotation speed" | Tweaked spec → clip | | Direct to Veo output | "Episode 2" / "Next scene" | End-to-end video clip render | Example prompts Full auto run > Make a 15-second narrative teaser about a rogue hacker finding a hidden code on the rooftop. Synthwave ambience, neon rain, slow roll-in. > Reference-driven > Image 1 = the knight, Image 2 = the sword. Turn into a 15-second duel scene on the rain-drenched battlements. > Series > Episode 1 of a 3-part micro-series. Cyberpunk courier discovers a glowing package. 15s cinematic. > When to use * Vertical aspect-ratio and short-form cinematic clips * Character-led moments with identity locked through cuts * Reference-heavy workflows—upload hero, location, prop, graphic, and poster * Micro-series or new thread—Each thread owns the tight story * Rapid concept runs—visual scene test, trailer wing * Turnkey ref-to-script, shots, references, standard, 15s duration Use the right tool when you need multi-branch grid, 1-per-second 16:9 editing, or multi-cut long-format videos without your own pipeline Output video_file_url 15s "Turn this into a video clip with consistent characters." Reference-driven—Shot-planned—One coherent clip.
Series
59 6 16
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/2966093/2026-10-05/9fdg96sq.jpg