10-of-my-most-popular-text-to-image-series-prompts-78b0897e
generate a bunch of images, then you curate the results to handpick the best ones
使用速推AI生成图像和视频。当用户需要画图、生成图片、制作视频、图生图、文生图、文生视频、图生视频时使用此 skill。支持即梦、Flux、Kling 等多种模型。
Follow the repository instructions for this skill and your agent. Review the files and dependencies before adding it to your workflow.
SkillHub CLI
skillhub install hexiaochun/sutui-aiInspect the source and compatibility requirements, then follow its installation steps.
Give your agent the source link and ask it to inspect instructions and dependencies before explaining the installation steps.
通过速推AI MCP服务进行AI图像和视频生成,支持多种模型。
user-速推AI
CallMcpTool 工具使用异步任务流程调用 Flux 2 Flash 模型:
// Step 1: 提交任务
{
"server": "user-速推AI",
"toolName": "submit_task",
"arguments": {
"model_id": "fal-ai/flux/dev",
"parameters": {
"prompt": "一只可爱的猫咪在花园里玩耍",
"image_size": "landscape_16_9"
}
}
}
// Step 2: 查询任务状态(轮询直到完成)
{
"server": "user-速推AI",
"toolName": "get_task",
"arguments": {
"task_id": "返回的任务ID"
}
}
备选同步方式(即梦模型,30-70秒):
{
"server": "user-速推AI",
"toolName": "sync_generate_image",
"arguments": {
"prompt": "一只可爱的猫咪在花园里玩耍",
"model": "jimeng-4.5"
}
}
图生图: 添加 image_url 参数即可
直接返回视频结果:
{
"server": "user-速推AI",
"toolName": "sync_generate_video",
"arguments": {
"prompt": "一只猫咪在奔跑,16:9",
"model": "jimeng-video-3.5-pro"
}
}
支持的模型: jimeng-video-3.5-pro、jimeng-video-3.5-pro-10s、jimeng-video-3.5-pro-12s
图生视频: 添加 image_url 参数
首尾帧视频: 同时添加 image_url 和 end_image_url
对于需要更多模型选择的场景,使用异步任务流程:
{
"server": "user-速推AI",
"toolName": "list_models",
"arguments": {
"category": "image" // 可选: image, video, audio, all
}
}
{
"server": "user-速推AI",
"toolName": "get_model_info",
"arguments": {
"model_id": "fal-ai/flux/schnell"
}
}
{
"server": "user-速推AI",
"toolName": "submit_task",
"arguments": {
"model_id": "fal-ai/flux/schnell",
"parameters": {
"prompt": "a beautiful sunset over the ocean",
"image_size": "landscape_16_9"
}...
These signals come from indexed source analysis and may lag behind changes. Check the current source files.
# 速推AI - 图像与视频生成
通过速推AI MCP服务进行AI图像和视频生成,支持多种模型。
## MCP 服务信息
- **服务标识**: `user-速推AI`
- **调用方式**: 使用 `CallMcpTool` 工具
## 快速开始
### 1. 生成图像(推荐 Flux 2 Flash)
使用异步任务流程调用 Flux 2 Flash 模型:
```json
// Step 1: 提交任务
{
"server": "user-速推AI",
"toolName": "submit_task",
"arguments": {
"model_id": "fal-ai/flux/dev",
"parameters": {
"prompt": "一只可爱的猫咪在花园里玩耍",
"image_size": "landscape_16_9"
}
}
}
// Step 2: 查询任务状态(轮询直到完成)
{
"server": "user-速推AI",
"toolName": "get_task",
"arguments": {
"task_id": "返回的任务ID"
}
}
```
**备选同步方式**(即梦模型,30-70秒):
```json
{
"server": "user-速推AI",
"toolName": "sync_generate_image",
"arguments": {
"prompt": "一只可爱的猫咪在花园里玩耍",
"model": "jimeng-4.5"
}
}
```
**图生图**: 添加 `image_url` 参数即可
### 2. 同步生成视频(推荐)
直接返回视频结果:
```json
{
"server": "user-速推AI",
"toolName": "sync_generate_video",
"arguments": {
"prompt": "一只猫咪在奔跑,16:9",
"model": "jimeng-video-3.5-pro"
}
}
```
**支持的模型**: jimeng-video-3.5-pro、jimeng-video-3.5-pro-10s、jimeng-video-3.5-pro-12s
**图生视频**: 添加 `image_url` 参数
**首尾帧视频**: 同时添加 `image_url` 和 `end_image_url`
## 异步任务流程
对于需要更多模型选择的场景,使用异步任务流程:
### Step 1: 查询可用模型
```json
{
"server": "user-速推AI",
"toolName": "list_models",
"arguments": {
"category": "image" // 可选: image, video, audio, all
}
}
```
### Step 2: 获取模型参数要求
```json
{
"server": "user-速推AI",
"toolName": "get_model_info",
"arguments": {
"model_id": "fal-ai/flux/schnell"
}
}
```
### Step 3: 提交任务
```json
{
"server": "user-速推AI",
"toolName": "submit_task",
"arguments": {
"model_id": "fal-ai/flux/schnell",
"parameters": {
"prompt": "a beautiful sunset over the ocean",
"image_size": "landscape_16_9"
}...generate a bunch of images, then you curate the results to handpick the best ones
generate a bunch of images, then you curate the results to handpick the best ones
Productivity system based on Maya calendar with 13 natural tones for project management and personal development
Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in (single or multi-account), or reading/injecting/running secrets via op.