# Seisoai

> AI image, video, music, audio, 3D, and LLM inference with x402 pay-per-request. Generate images with FLUX/FLUX-2, videos with Veo 3.1/LTX-2, music with CassetteAI, voice clone, lip sync, transcription, and Claude LLM inference. Create custom AI agents. Use when the user wants AI generation, creative content, chat completions, or agentic workflows.

- Skill: `kernel8901/seisoai` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add kernel8901/seisoai`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kernel8901/seisoai/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: kernel8901 (https://skillmd.com/u/kernel8901)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kernel8901/seisoai

---


# Seisoai

Generate AI images, videos, music, audio, 3D models, and run LLM inference. Pay per request with USDC on Base — no account needed.

**Base URL:** `https://seisoai.com`

---

## Quick Start (Recommended for Agents)

The fastest way to use Seisoai is through the **Gateway API**. One endpoint for everything.

### 1. Discover Available Tools

```
GET /api/gateway/tools
```

Returns all tools with pricing and input schemas.

### 2. Invoke Any Tool

```
POST /api/gateway/invoke/{toolId}
Content-Type: application/json

{
  "prompt": "a sunset over mountains",
  ...tool-specific params
}
```

### 3. Handle x402 Payment

First request returns `HTTP 402`. Decode `PAYMENT-REQUIRED` header, sign payment, retry with `PAYMENT-SIGNATURE` header.

### Common Tool IDs

| Task | Tool ID | Price |
|------|---------|-------|
| Image (fast) | `image.generate.flux-pro-kontext` | $0.065 |
| Image (photorealistic) | `image.generate.flux-2` | $0.0325 |
| Image (premium) | `image.generate.nano-banana-pro` | $0.325 |
| Video (quality) | `video.generate.veo3` | $0.13/sec |
| Video (fast) | `video.generate.ltx-text` | $0.052/sec |
| Music | `music.generate` | $0.026/min |
| Voice clone | `audio.tts` | $0.026 |
| Transcribe | `audio.transcribe` | $0.013 |
| Image to 3D | `3d.image-to-3d` | $0.195 |

---

## When to Use Which Tool

```
User wants an image?
├─ Photorealistic / text in image → flux-2
├─ Fast general purpose → flux-pro-kontext  
├─ 360° panorama → nano-banana-pro
└─ Edit existing image → flux-pro-kontext-edit

User wants a video?
├─ High quality / complex scene → veo3
├─ Fast / simple scene → ltx-text
├─ Animate an image → veo3-image-to-video or ltx-image
└─ First + last frame → veo3-first-last-frame

User wants audio?
├─ Music from description → music.generate
├─ Sound effect → audio.sfx
├─ Clone voice / TTS → audio.tts
├─ Transcribe speech → audio.transcribe
└─ Separate stems → audio.stem-separation

User wants to edit an image?
├─ Swap faces → image.face-swap
├─ Fill in area → image.inpaint
├─ Extend image → image.outpaint
├─ Remove background → image.extract-layer
└─ Upscale → image.upscale

User wants 3D?
└─ Image to GLB model → 3d.image-to-3d

User wants multi-step workflow?
└─ Use /api/gateway/orchestrate with goal description
```

---

## x402 Payment Flow

All endpoints support x402 pay-per-request. No account needed.

```
1. POST /api/gateway/invoke/image.generate.flux-2
   → 402 Payment Required + PAYMENT-REQUIRED header

2. Decode header (base64 JSON):
   {
     "accepts": [{
       "scheme": "exact",
       "network": "eip155:8453",
       "maxAmountRequired": "32500",
       "asset": "USDC",
       "payTo": "0x..."
     }]
   }

3. Sign USDC payment with wallet

4. Retry with header:
   PAYMENT-SIGNATURE: <signed payload>

5. Success response includes:
   { "x402": { "settled": true, "transactionHash": "0x..." } }
```

### User Confirmation Required

Before signing any payment, agents MUST:
- Show amount (e.g., "$0.065 USDC")
- Show what's being purchased
- Get explicit user approval

```
Agent: Generate image of sunset? Cost: $0.065 USDC on Base. [Confirm/Cancel]
User: Confirm
Agent: [signs, generates, returns image]
```

---

## Core Endpoints

### Image Generation

**POST /api/generate/image**

```json
{
  "prompt": "a sunset over mountains, oil painting style",
  "model": "flux-pro",
  "aspect_ratio": "16:9",
  "num_images": 1
}
```

| Param | Type | Default | Description |
|-------|------|---------|-------------|
| `prompt` | string | required | Image description |
| `model` | string | `flux-pro` | `flux-pro`, `flux-2`, `nano-banana-pro` |
| `aspect_ratio` | string | `1:1` | `1:1`, `16:9`, `9:16`, `4:3`, `3:4` |
| `num_images` | int | 1 | 1-4 |
| `image_url` | string | - | Reference image for editing |
| `seed` | int | - | Reproducibility seed |

**Response:**
```json
{
  "success": true,
  "images": ["https://fal.media/files/..."],
  "x402": { "settled": true, "transactionHash": "0x..." }
}
```

---

### Video Generation

**POST /api/generate/video**

```json
{
  "prompt": "a cat walking through a garden",
  "model": "veo",
  "duration": "6s",
  "generate_audio": true
}
```

| Param | Type | Default | Description |
|-------|------|---------|-------------|
| `prompt` | string | required | Video description |
| `model` | string | `veo` | `veo` (quality) or `ltx` (fast) |
| `duration` | string | `6s` | `4s`, `6s`, `8s` for Veo; 1-10s for LTX |
| `generate_audio` | bool | true | Generate synchronized audio |
| `first_frame_url` | string | - | Starting frame image |
| `last_frame_url` | string | - | Ending frame image |

**Response:**
```json
{
  "success": true,
  "video": { "url": "https://fal.media/files/...", "content_type": "video/mp4" },
  "x402": { "settled": true, "transactionHash": "0x..." }
}
```

---

### Music Generation

**POST /api/generate/music**

```json
{
  "prompt": "upbeat jazz with piano and drums, 120 BPM",
  "duration": 60,
  "selectedGenre": "jazz"
}
```

| Param | Type | Default | Description |
|-------|------|---------|-------------|
| `prompt` | string | required | Music description (include genre, instruments, mood, tempo) |
| `duration` | int | 60 | Duration in seconds (10-180) |
| `selectedGenre` | string | - | `lo-fi`, `electronic`, `orchestral`, `rock`, `jazz` |

---

### Voice Clone / TTS

**POST /api/audio/voice-clone**

```json
{
  "text": "Hello, this is a test of voice cloning.",
  "voice_url": "https://example.com/reference-voice.wav",
  "language": "en"
}
```

---

### Speech-to-Text

**POST /api/audio/transcribe**

```json
{
  "audio_url": "https://example.com/speech.mp3",
  "task": "transcribe"
}
```

| Param | Type | Default | Description |
|-------|------|---------|-------------|
| `audio_url` | string | required | Audio/video URL |
| `task` | string | `transcribe` | `transcribe` or `translate` (to English) |
| `chunk_level` | string | `segment` | `segment` or `word` for timestamps |

---

### Sound Effects

**POST /api/audio/sfx**

```json
{
  "prompt": "thunder rumbling in the distance",
  "duration": 5
}
```

---

### Image to 3D

**POST /api/model3d/generate**

```json
{
  "image_url": "https://example.com/object.jpg",
  "generate_type": "Normal"
}
```

| Type | Price | Output |
|------|-------|--------|
| `Normal` | $0.195 | Full textured GLB |
| `Geometry` | $0.13 | Geometry only |
| `LowPoly` | $0.104 | Low-poly mesh |

---

## Image Tools

### Face Swap
```
POST /api/image-tools/face-swap
{ "source_image_url": "...", "target_image_url": "..." }
Price: $0.026
```

### Inpainting
```
POST /api/image-tools/inpaint
{ "image_url": "...", "mask_url": "...", "prompt": "a red sports car" }
Price: $0.026
```

### Outpainting
```
POST /api/image-tools/outpaint
{ "image_url": "...", "prompt": "extend the landscape", "direction": "all" }
Price: $0.026
```

### Background Removal
```
POST /api/extract/layers
{ "image_url": "..." }
Price: $0.039
```

### Upscale
```
POST /api/generate/upscale
{ "image_url": "..." }
Price: $0.039
```

### Describe Image
```
POST /api/image-tools/describe
{ "image_url": "...", "detail_level": "detailed" }
Price: $0.013
```

---

## Multi-Step Orchestration

Let Seisoai plan and execute complex workflows automatically.

**POST /api/gateway/orchestrate**

```json
{
  "goal": "Create a music video: generate an image of a sunset, animate it to video, add jazz music",
  "planOnly": false
}
```

The orchestrator will:
1. Generate the sunset image
2. Convert to video with Veo
3. Generate jazz music
4. Combine audio + video

**Response includes step-by-step results with all generated assets.**

---

## Pre-built Workflows

| Workflow | Endpoint | Description | Est. Cost |
|----------|----------|-------------|-----------|
| AI Influencer | `POST /api/workflows/ai-influencer` | Portrait → Voice → Lip Sync | ~$0.15 |
| Music Video | `POST /api/workflows/music-video` | Image → Video → Music | ~$1.00 |
| Avatar Creator | `POST /api/workflows/avatar-creator` | Description → Variations | ~$0.20 |
| Stem Separator | `POST /api/workflows/remix-visualizer` | Audio → Separated Stems | ~$0.05 |

---

## LLM Chat

**POST /api/chat-assistant/message**

```json
{
  "message": "Help me create a prompt for a fantasy landscape",
  "model": "claude-sonnet-4-5"
}
```

| Model | Best for |
|-------|----------|
| `claude-haiku-4-5` | Fast, simple tasks |
| `claude-sonnet-4-5` | Balanced (default) |
| `claude-opus-4-6` | Complex reasoning |

---

## Pricing Summary

| Category | Tool | Price |
|----------|------|-------|
| **Image** | FLUX Pro | $0.065 |
| | FLUX-2 | $0.0325 |
| | Nano Banana Pro | $0.325 |
| **Video** | Veo 3.1 | $0.13/sec |
| | LTX-2 | $0.052/sec |
| **Audio** | Music | $0.026/min |
| | SFX | $0.039 |
| | Voice Clone | $0.026 |
| | Transcribe | $0.013 |
| **Image Tools** | Face Swap | $0.026 |
| | Inpaint/Outpaint | $0.026 |
| | Upscale | $0.039 |
| | Background Remove | $0.039 |
| | Describe | $0.013 |
| **3D** | Image to 3D | $0.13-$0.195 |
| **Training** | LoRA | $0.0026/step |

All prices are final (include 30% markup over API costs).

---

## Error Handling

| Code | Meaning | Action |
|------|---------|--------|
| 402 | Payment required | Decode `PAYMENT-REQUIRED` header, sign, retry |
| 400 | Bad request | Check required params in error message |
| 500 | Server error | Retry or report issue |

---

## For Agent Developers

### Minimal Integration

```python
import requests
import base64
import json

BASE = "https://seisoai.com"

def generate_image(prompt, wallet_sign_fn):
    # First request gets 402
    r = requests.post(f"{BASE}/api/gateway/invoke/image.generate.flux-2", 
                      json={"prompt": prompt})
    
    if r.status_code == 402:
        # Decode payment requirements
        payment_req = json.loads(base64.b64decode(r.headers["PAYMENT-REQUIRED"]))
        amount = payment_req["accepts"][0]["maxAmountRequired"]
        
        # Get user confirmation
        if not confirm_with_user(f"Generate image for ${int(amount)/1e6:.4f} USDC?"):
            return None
        
        # Sign and retry
        signature = wallet_sign_fn(payment_req)
        r = requests.post(f"{BASE}/api/gateway/invoke/image.generate.flux-2",
                          json={"prompt": prompt},
                          headers={"PAYMENT-SIGNATURE": signature})
    
    return r.json()["images"][0] if r.ok else None
```

### Best Practices

1. **Use Gateway API** - One interface for all tools
2. **Cache tool list** - `GET /api/gateway/tools` doesn't change often
3. **Handle 402 gracefully** - It's expected, not an error
4. **Always confirm payments** - Never auto-sign without user approval
5. **Use orchestrate for complex tasks** - Let Seisoai plan multi-step workflows

---

## Config

```json
{
  "skills": {
    "entries": {
      "seisoai": {
        "enabled": true,
        "config": {
          "apiUrl": "https://seisoai.com"
        }
      }
    }
  }
}
```