Story Studio Short-Video Pipeline

Helps users make historical narrative and real-person story short videos in Chinese, going from topic selection all the way to a draft that can be further edited in CapCut. The underlying project is the open-source story-studio (author konlu, license CC BY-NC-SA 4.0, non-commercial use only).

Full workflow: topic → narration script → AI voice-over → subtitle alignment → storyboard illustrations → CapCut draft. You can run the whole thing, or just one part of it.

You execute the process: call story-studio for the scope the user asks for, and do not do extra work or spend money on the user's behalf.

Triggers

The user mentions story-studio or 故事工坊, or wants to make a Chinese short video about history or real-person stories: find a topic, write a narration script, revise the opening hook, do voice-over, subtitles, split into storyboards, illustrate, or export a CapCut draft. Also use it when the user already has a script or voice-over and only wants the later steps.

Not suitable: topics that are neither historical nor about people (technology, food, etc.), since the built-in writing styles and art styles are aimed at the former.

Hard Rules

1. The license is non-commercial. This project only allows personal learning and non-commercial creation. Do not help the user resell it, use it for paid courses, or offer it as a paid service; if the user has commercial use, tell them to contact the author first.
2. Confirm before spending money. Voice synthesis (Volcano Engine Doubao speech) and image generation may incur costs. Before the first synthesis or any batch image generation, tell the user what will be done and roughly how many times, and wait for confirmation. Do not re-synthesize the same script with the same voice.
3. API keys stay out of the conversation. The user sets the Doubao speech API Key as the environment variable DOUBAO_TTS_API_KEY themselves; do not ask them to paste it to you, and do not print it or write it into project files.
4. Historical accuracy first. List facts and timeline first, distinguishing established history, disputed accounts, and dramatized fiction. Do not invent numbers, quotes, or suffering details; anything unverified is marked as pending verification.
5. Only do the requested stages. If the user says "just write the script," do not also do voice-over or images. If materials are missing, report the gap rather than quietly enabling other stages.
6. Put outputs in the project folder, not in the skill directory.

Installation

First check whether it is already installed:

ls ~/.claude/skills/story-studio 2>/dev/null && echo "installed" || echo "not installed"

If not installed, clone it into the skills directory:

git clone https://github.com/konlu/story-studio.git ~/.claude/skills/story-studio

After installing, restart Claude Code once. The first time voice-over, transcription, or draft export is needed, let story-studio check and configure the runtime itself (it will prepare Python, Node.js, FFmpeg, the local subtitle recognition tool Coli, and the CapCut draft library in the user data directory). If you only write scripts, none of these are installed.

CapCut Pro: the repository notes say only Windows can locate the draft directory with one click; for other systems, follow the repository's references/setup.md.

First Ask Which Segments to Run

story-studio uses three switches to set the scope. If it is not clear, ask one question; do not default everything to on:

User wants to do	Writing	Illustrations	CapCut draft
Find topics, write narration script, revise hook	On	Off	Off
Script already exists, split storyboards and illustrate	Off	On	Off
Voice-over and images exist, only arrange CapCut draft	Off	Off	On
Make a complete episode from scratch	On	On	On

Also ask about these settings: writing style (historical narrative / real-person story), target word count, illustration art style (thick-paint historical comic / black-and-white documentary / historical realism), and whether voice-over uses Doubao synthesis or a user-recorded MP3.

Common Requests

Full episode production (replace the brackets with the user's content):

Use story-studio to make a short video about [person or event].
Writing: [historical narrative / personal story] style, about [word count] words. Give me 3 openings with different angles to choose from, and continue only after the script is final.
Voice-over: [synthesize it with Doubao / I'll record it myself and put it in the folder].
Images: after aligning to the timeline, split into [number] shots, [art style], 9:16 vertical.
Draft: export a CapCut draft with slow push-in and pull-out camera moves, no subtitles yet.
Ask me before any step that costs money.

Script only, with fact checking:

Use story-studio to write only a narration script about [person], about [word count] words.
Hand in a fact-check table and a timeline first, then the script. Script only: no voice-over, no images.

Existing script and voice-over, continue the rest:

Continue the project in the current folder: the final script is script.md and the voice-over is voice.mp3.
Align the subtitles with local tools, split it into shots and add images, then export a CapCut draft.

Only storyboards and prompts, images generated by another tool:

Only split it into shots; don't call any image-generation tool.
Read the final script and voice-over in the folder, and output a storyboard table with timestamps plus a complete image prompt for each shot.

Post-Production Checks

• The first sentence has matching visuals from second 0; each image corresponds to the person and event being discussed at that moment in the narration.
• Names of people, places, years, and numbers in the script can all be found in the fact table.
• The draft opens in CapCut, and images and voice-over have been copied into the draft directory.
• Costs: state how many voice-over and image generation calls were actually used this time.

Output Format

1. First state which segments were completed and which were not.
2. List the output paths (script, voice-over, subtitles, storyboards, CapCut draft).
3. List items pending user confirmation or supplement: historical facts to verify, missing materials, costs not yet incurred.
4. Final sentence reminder: this project is for non-commercial use only, and the user should double-check the history themselves before publishing.