Learning Timeline
Key Insights
Claude's Advantages for Prompt Engineering
Claude (especially the Opus model) excels at 'prompt engineering' for vision compared to other models. It is capable of processing complex inputs to generate highly specific instructions.
Prompt Strategy Differences
Unlike models like Kling 3 which are more effective with simple prompts, Seedance requires highly detailed and specific prompts to achieve 'crushed it' or extraordinary video quality.
The Importance of Reference Images
Use a reference image as the 'source of truth'. The AI will mimic the art style and visual taste of that image, acting as a personal assistant that understands your artistic vision.
Prompts
Vision Prompt Optimization for Claude
Target:
Claude Opus
I am using Seedance to create a video. Based on this reference image and my idea [INSERT IDEA], please write a highly detailed and specific video prompt. Focus on preserving the character identity, describing the exact motions, and detailing the transitions to ensure a high-quality output.
Step by Step
Seedance Video Prompt Optimization Workflow
- Prepare a high-quality 'Source Reference Image' that represents your artistic vision.
- Open your web browser and access the Claude platform (Opus model recommended).
- Upload the reference image to Claude so the AI can understand your visual style and 'taste'.
- Write a brief description of your original vision or idea in the Claude chat.
- Instruct Claude to generate a 'highly specific' prompt focusing on 'character identity', motions, and video transitions.
- Copy the optimized prompt generated by Claude.
- Open the Seedance platform (version 2.0 or latest).
- Upload the same reference image into the 'Reference Image' slot in Seedance.
- Paste the optimized prompt into the Seedance prompt input box.
- Click the 'Generate' button to produce a high-quality video output.
