multimodal-gen

Multimodal Content Generation

When the user needs to generate images or videos, automatically call multimodal-agent for processing.

Trigger Scenarios

Image Generation

• "Generate an image"
• "Draw a..."
• "AI drawing"
• "Text-to-image"
• "Help me generate an image"
• "Generate with flux/imagen/dalle"

Video Generation

• "Generate a video"
• "Make a video"
• "Text-to-video"
• "Generate a video with veo/sora/kling"

Usage

Automatically call multimodal-agent

sessions_spawn(
    agentId="multimodal-agent",
    task="生成图片: {用户描述}, 使用 {模型} 模型"
)

Available Models

Image Generation

Alias	Model	Features
flux	flux-pro-max	High quality, recommended
imagen	google/imagen-4-ultra	Google's strongest
dalle	gpt-image-1	DALL-E 3
doubao	doubao-seedream-4-5	Chinese aesthetics
klingimg	kling-image	Kling image generation

Video Generation

Alias	Model	Features
veopro	veo3.1-pro	Google Professional Edition
veo4k	veo3.1-pro-4k	4K HD
sora	sora-2-pro-all	OpenAI Sora
kling	kling-video	Kling video

Execution Flow

用户请求 "生成一张猫咪图片"
    │
    ▼
┌─────────────────────────────────────┐
│ 1. 识别为图片生成请求              │
│ 2. 提取描述: "猫咪"                │
│ 3. 选择默认模型: flux-pro-max      │
└─────────────────────────────────────┘
    │
    ▼
┌─────────────────────────────────────┐
│ sessions_spawn(                     │
│   agentId="multimodal-agent",       │
│   task="生成图片: 一只可爱的猫咪,   │
│         使用 flux 模型"             │
│ )                                   │
└─────────────────────────────────────┘
    │
    ▼
multimodal-agent 执行生成并返回结果

Examples

Generate Image

用户: 帮我生成一张日落海滩的图片

执行:
sessions_spawn(
    agentId="multimodal-agent",
    task="生成图片: 日落时分的海滩，金色阳光洒在海面上，使用 flux 模型"
)

Generate Video

用户: 用 sora 生成一个猫咪玩耍的视频

执行:
sessions_spawn(
    agentId="multimodal-agent",
    task="生成视频: 一只可爱的猫咪在草地上玩耍，使用 sora 模型"
)

Specify Model

用户: 用 doubao 画一张中国风山水画

执行:
sessions_spawn(
    agentId="multimodal-agent",
    task="生成图片: 中国风山水画，云雾缭绕的山峰，使用 doubao 模型"
)

Model Selection Recommendations

Scenario	Recommended Model
General high quality	flux
Chinese style	doubao
Realistic photo	imagen
Creative art	dalle
HD video	veo4k
Cinematic video	sora
Fast video	kling

Notes

1. Prompt optimization: multimodal-agent will automatically optimize the user's description
2. Model selection: If the user does not specify, use flux (image) or veopro (video) by default
3. Asynchronous execution: Video generation may take a long time and will run in the background
4. Result return: After generation is complete, the result will be automatically sent to the user