explaingit

zhouwei713/seedance-prompt

Analysis updated 2026-05-18

257Audience · vibe coderComplexity · 1/5LicenseSetup · easy

TLDR

An AI agent skill that writes realistic AI video prompts by first deciding who filmed the footage, on what device, and what flaws that would leave behind.

Mindmap

mindmap
  root((seedance-prompt))
    What it does
      Realistic video prompts
      Device flaw reference
      Anti AI checklist
    Tech stack
      Markdown skill file
    Use cases
      AI video generation
      Prompt rewriting
      Style reference
    Audience
      Vibe coders
      Creative prompters
    License
      MIT

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Turn a short scene idea into a detailed, realistic prompt for AI video tools like Seedance or Sora.

USE CASE 2

Add device specific camera flaws, like VHS artifacts or smartphone shake, to make generated video feel less polished and more authentic.

USE CASE 3

Use the built in checklist to catch overly clean, AI looking output before generating a video.

USE CASE 4

Install the skill into Hermes, Claude, Codex, or Cursor to reuse the prompting method across projects.

What is it built with?

Markdown

How does it compare?

zhouwei713/seedance-prompteasylink-ai-open/tiri-agentkitlangton/skills
Stars257257257
LanguageTypeScript
Setup difficultyeasymoderateeasy
Complexity1/54/51/5
Audiencevibe coderdeveloperdeveloper

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · easy Time to first run · 5min

Reference material and examples are written in Chinese.

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

In plain English

This project is an Agent Skill, a reusable instruction file, for writing more realistic prompts when generating AI video with tools like Seedance, Sora, Kling, Runway, or Veo. Its main idea is that most AI video prompts describe a pretty picture but skip a key question: who supposedly filmed this footage, on what device, and in what era. The README explains that answering those questions first, then writing in the resulting camera flaws, natural mistakes, ambient sound, and a timeline, is what makes generated video feel like it actually happened rather than looking like a polished AI sample. The README gives two worked examples in Chinese. One turns the phrase morning on a New York street into a prompt describing a smartphone video supposedly shot vertically by a passerby, including a consistent character description, a timed shot list, and camera quirks like exposure shifts and slight shake. The other turns a short description of an old Korean neighborhood into a prompt styled after a 2000s consumer camcorder, complete with compression artifacts, faded color, and small unplanned moments like a cat wandering into frame. The skill breaks a video idea into seven parts: main character, location, visual style, camera style, timeline, audio, and overall goal, and it includes built in reference material covering device specific visual flaws for things like VHS camcorders, film cameras, smartphones, and security cameras, a dictionary that translates moods like warmth or loneliness into concrete filmable details, and a checklist for catching leftover AI looking polish before finishing a prompt. You install it by cloning the repository into a Hermes skills folder, or by dropping its SKILL.md file into other AI agent tools such as Claude, Codex, Cursor, or OpenCode that support skills or long prompts, or you can simply read the included reference files directly as a prompting guide. The project is released under the MIT license, which allows free use, modification, and commercial use as long as you keep the copyright notice.

Copy-paste prompts

Prompt 1
Turn this scene idea into a realistic AI video prompt using this skill's seven part structure.
Prompt 2
Rewrite my video prompt to remove overly clean, AI looking camera movement and color.
Prompt 3
Give this prompt a 2000s consumer camcorder look using this repo's device flaw reference.
Prompt 4
Check my video prompt against this repo's anti AI look checklist.

Frequently asked questions

What is seedance-prompt?

An AI agent skill that writes realistic AI video prompts by first deciding who filmed the footage, on what device, and what flaws that would leave behind.

What license does seedance-prompt use?

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

How hard is seedance-prompt to set up?

Setup difficulty is rated easy, with roughly 5min to a first successful run.

Who is seedance-prompt for?

Mainly vibe coder.

Open on GitHub → Explain another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.