按模态浏览 PROMPTS

视频 Prompts — 38 个精选示例

面向 视频 生成工作流的 prompt 示例。按分类筛选,复制 prompt,或打开模型专属 API 代码。

VI
视频
social-media hailuo-02-text-to-video-pro

Social media — coffee pour in slow motion

Close-up slow-motion shot of espresso being poured into a clear glass containing cold milk and ice cubes. The dark espresso hits the milk surface and creates mesmerizing turbulent swirls and marbling patterns as the two liquids mix. Shot from a slightly elevated angle looking down into the glass. The espresso stream is thin and steady, pouring from just above frame. Condensation beads on the outside of the glass. Background is a soft blurred café counter with warm bokeh light spots. High frame rate slow motion capturing every tendril of mixing liquid. Vertical 9:16 format optimized for Instagram Reels.

查看 API 代码
curl -X POST https://runapi.ai/api/v1/hailuo/text_to_video \
  -H "Authorization: Bearer $RUNAPI_KEY" \
  -H "Content-Type: application/json" \
  --data-binary @- <<'JSON'
{
  "model": "hailuo-02-text-to-video-pro",
  "prompt": "Close-up slow-motion shot of espresso being poured into a clear glass containing cold milk and ice cubes. The dark espresso hits the milk surface and creates mesmerizing turbulent swirls and marbling patterns as the two liquids mix. Shot from a slightly elevated angle looking down into the glass. The espresso stream is thin and steady, pouring from just above frame. Condensation beads on the outside of the glass. Background is a soft blurred café counter with warm bokeh light spots. High frame rate slow motion capturing every tendril of mixing liquid. Vertical 9:16 format optimized for Instagram Reels."
}
JSON
常见问题

使用视频 prompts

好的%{model} prompt 应该包含什么?

有用的%{model} prompt 会清楚写出主体、风格、约束和输出目标。这里的示例足够短,便于复制;同时也足够具体,agent 或后端任务能保留生成意图。

这些 prompts 可以跨模型复用吗?

多数情况下可以。先从同一模态的 prompt 开始,再按模型调整画幅、时长、声音设置或风格控制。详情页会把保存过的参数列在 prompt 文本旁边。

在哪里找正确的模型 slug?

每张卡片都会显示 RunAPI 模型 slug。只想看一个模型的示例时,打开模型页;需要定价和能力细节时,进入模型目录页查看。

agent 可以直接调用这些 prompts 吗?

可以。安装 RunAPI MCP Server 后,复制 prompt 详情页里的 agent 指令。页面会把 prompt 文本、模型 slug 和 endpoint 路径放在一起,给 agent 足够上下文。