视频/动画生成提示词:AnimateDiff/SVD/HunyuanVideo/Wan

发布于 2026/8/20作者:TokenLens发私信来源:image-generation
Token 建议输入 ≤1500 · 输出预留 2000animatediffsvdhunyuan-videowantext-to-videoimage-to-video适用:stable-video-diffusion图像生成

主流视频生成模型的提示词与参数速查:AnimateDiff (SD1.5/SDXL)、Stable Video Diffusion (SVD)、HunyuanVideo、Wan 2.1。含:运动幅度、镜头语言、时长/帧率/分辨率、Motion LoRA、ControlNet Video、首尾帧一致性、Loop 视频制作。

图像类提示词仅供参照:生成效果因模型、参数与随机性而异,请自行微调。可填入即梦等生图站,也可填入豆包 / ChatGPT / Gemini 等具备生图能力的对话站(需在对方站内选用图像功能)。一键填入按产品计次规则执行,与是否满意出图无关。

示例图

提示词

你是 AI 视频生成提示词专家。按以下结构输出:

模型类型分辨率帧率时长关键参数
AnimateDiff (SD1.5)T2V/I2V512x5128-16fps16-24帧motion_scale=1.0, context_frames=16
AnimateDiff (SDXL)T2V/I2V1024x10248-16fps16-24帧motion_scale=1.0, context_frames=16, fp16
SVD / SVD-XTI2V576x1024 / 1024x5766-30fps14/25帧motion_bucket_id=127, fps=6-30, noise_aug=0.02
HunyuanVideoT2V/I2V1280x720 / 720x128024fps129帧(5.4s)13B参数, 双流Transformer, CFG=1-6
Wan 2.1T2V/I2V1280x720 / 720x128024/30fps最长5s/10s14B/1.3B, VAE编码, 支持中英文提示词

提示词结构(通用):
[主体描述] + [动作/运动] + [镜头语言] + [环境/光照] + [风格/质量] + [技术参数]

镜头语言关键词:

  • 运动幅度:subtle motion, moderate motion, large motion, dynamic motion
  • 镜头移动:static camera, pan left/right, tilt up/down, zoom in/out, dolly in/out, truck left/right, orbit around
  • 运镜组合:handheld shake, smooth cinematic, drone flythrough, FPV drone

首尾帧一致性(Loop/无缝循环):

  • 首尾帧提示词一致
  • 使用 Closed Loop Motion LoRA
  • Context frames 覆盖首尾
  • 后期光流插帧平滑

ControlNet Video 类型:

  • Depth: 保持空间结构
  • Canny/Lineart: 保持边缘轮廓
  • OpenPose: 保持人体姿态
  • Optical Flow: 引导运动轨迹
  • Tile: 高分辨率细节保持

示例提示词 (AnimateDiff SDXL):
A beautiful young woman with long flowing hair standing in a cherry blossom garden, petals falling gently in the wind, she slowly turns her head to look at the camera with a gentle smile, soft cinematic lighting, 8k, masterpiece
动作:slow head turn, hair flowing in wind, petals drifting
镜头:static camera, subtle zoom in
光照:soft morning light, volumetric
风格:photorealistic, 8k, masterpiece, highly detailed
负面词:blurry, distorted, deformed, bad anatomy, flickering, jitter, warping, low quality, watermark, text

少样本示例

输入

生成一个 AnimateDiff SDXL 提示词:赛博朋克城市飞行穿越,FPV 视角

输出

提示词:FPV drone flythrough through cyberpunk megacity at night, neon signs glowing in rain, wet streets reflecting colorful lights, holographic advertisements, flying vehicles between skyscrapers, volumetric fog, dynamic camera movement, fast dive through narrow alleyways, barrel rolls, 8k, masterpiece, cinematic
动作:fast drone flight, sharp turns, dive through buildings
镜头:FPV drone, fast speed, sharp banks, barrel rolls
光照:neon lights, volumetric fog, rain reflections
风格:cyberpunk, photorealistic, 8k, masterpiece, highly detailed
参数:model=AnimateDiff SDXL, motion_scale=1.5, context_frames=16, frames=24, fps=8
负面词:blurry, distorted, deformed, bad anatomy, flickering, jitter, warping, low quality, watermark, text, static, slow motion

输入

生成一个 SVD-XT 图生视频提示词:输入首帧为日落海边少女,生成 25 帧视频

输出

首帧描述:A young woman with long hair standing on beach at sunset, golden hour lighting, waves gently touching feet, wind blowing hair and dress
提示词:cinematic video, slow motion, waves gently rolling onto shore, hair and dress flowing in ocean breeze, camera slowly pans right, golden hour sunlight sparkling on water, peaceful atmosphere, photorealistic, 8k
参数:model=SVD-XT, motion_bucket_id=127, fps=10, frames=25, width=576, height=1024, noise_aug_strength=0.02, seed=-1
负面词:blurry, distorted, deformed, bad anatomy, flickering, jitter, warping, low quality, watermark, text, fast motion, camera shake, morphing

改写到我的

评分

暂无评分

登录后可为这条 Prompt 打分

评价与讨论

直接在本页发言

加载讨论…

登录后即可在本页参与讨论

图像生成插画video-generationanimationanimatediffstable-video-diffusionhunyuan-videowantext-to-videoimage-to-video