zhouwei713/seedance-prompt

What it solves

AI-generated videos often look like overly polished commercials or generic demos because they are too stable, too clean, and lack the natural imperfections of real-world footage. This project provides a structured framework to move beyond generic descriptors like "cinematic" or "8K" and instead design prompts that simulate real-world recording equipment, human error, and authentic environmental details.

How it works

The system transforms a simple video idea into a detailed, structured prompt divided into seven key components:

1. Main Characters: Ensures consistent appearance and clothing across shots.
2. Location: Defines specific spatial, temporal, and environmental details.
3. Visual Style: Sets the overall mood and material quality.
4. Camera Style: Specifies the device (e.g., 2000s DV, VHS, smartphone), including intentional flaws like motion blur, autofocus hunting, or exposure fluctuations.
5. Timeline: Maps out a 10-15 second sequence of events, including "non-perfect" moments.
6. Audio: Matches ambient sounds to the visual action.
7. Goal: Defines the final intended identity of the footage.

It utilizes reference libraries for equipment aesthetics (e.g., Super 8 film, CCTV), an atmosphere dictionary to translate abstract moods into visual phenomena, and an "anti-AI" checklist to remove commercial-style polish.

Who it’s for

Creators using text-to-video or image-to-video models such as Seedance, Sora, Kling, Runway, and Veo who want to achieve a documentary, home-movie, or candid street-photography aesthetic and avoid the "AI look."

Highlights

• Equipment Simulation: Mimics specific hardware flaws like scan lines, digital artifacts, and handheld shake based on the chosen era/device.
• Candid Logic: Focuses on life-like movements and unplanned events rather than staged poses.
• Structured Timeline: Breaks down the video into a second-by-second sequence to maintain narrative flow and character consistency.
• Anti-AI Filtering: Specifically identifies and removes overly stable camera movements and commercial color grading.