A prompt that produces a great result on one model can fall flat on another — each model reads pacing, detail level, and motion cues differently. Rather than one generic prompt-writing guide, this breaks it down by the model families that matter most, with real working examples for each linked below.
Start generating freeWan responds well to direct, concrete prompts without excessive layering — it's the cheapest model to iterate on, so the practical approach is writing a simple version first and refining based on the result rather than front-loading every detail. See real working examples at /prompts/wan-ai-prompts.
Seedance is tuned for cinematic camera motion, so a prompt that only describes a static scene wastes that capability — include camera direction and shot progression explicitly. Real working examples: /prompts/seedance-ai-prompts.
Seedream rewards prompts that spell out material, pose, and light direction rather than a vague one-line description — the more specific the prompt, the sharper and more accurate the result. Real working examples: /prompts/seedream-ai-prompts.
A generic prompt on an anime-capable model tends to drift toward photoreal shading within a few words — explicitly naming cel shading, clean linework, and anime proportions in the prompt keeps the style locked in. Real working examples: /prompts/anime-nsfw-ai-prompts.
Not reliably — each model reads pacing, detail level, and motion cues differently. A prompt tuned for Seedream's level of detail may underperform on Wan, and vice versa.
Model-specific prompt libraries are linked in each section above: /prompts/wan-ai-prompts, /prompts/seedance-ai-prompts, /prompts/seedream-ai-prompts, and /prompts/anime-nsfw-ai-prompts.
No — every example prompt describes a fictional, AI-generated adult character, always depicted as an adult. Prompts describing real, identifiable people or anyone under 18 are banned by the content rules.











