# 嘿鲸 (Hey Whale) — AI creation studio by MrWhale

> **Humans:** https://hey.mrwhale.ai — web studio for image, writing, comics, style templates, and short video.
> **Agents:** x402 pay-per-call GPT Image 2 API below (no login / no API key).

## Product (web studio)
嘿鲸 is the creator-facing studio on MrWhale.

| Tool | Role |
|------|------|
| 鲸图 | Text / reference → images |
| 鲸笔 | Idea → outline, chapters, prose |
| 鲸漫 | Story → comic pages |
| 鲸仿 | Image → fill-in style template |
| 鲸影 | Image → short video |
| 画布 / 素材 / 作品 | Refine, reusable assets, library |

Site: https://hey.mrwhale.ai
Brand: https://mrwhale.ai/

## Agent Image API (x402)

> Engine = **GPT Image 2**. Pay per call with x402 (USDC on Base).
> Domestic human billing (epay/quota) is the web path — not this Agent path.

### Quick facts
- Product: 嘿鲸 Agent Image API
- Engine: GPT Image 2
- Gateway model id: `jingtu-image-1k` (aliases: gpt-image-2, GPT Image 2, jingtu-image-1k)
- Price: $0.01 USDC per 1K image (pay-per-call, n=1)
- Network: base (eip155:8453)
- Protocol: x402 exact scheme
- Pay to: 0x68a361264F0c8e93242cD36A0C1992fC687d9B8a
- Output: PNG URL (async job)

### Endpoints
- Generate: POST https://hey.mrwhale.ai/api/agent/v1/images/generations
- Job status: GET https://hey.mrwhale.ai/api/agent/v1/images/jobs/{id}
- Pricing + params JSON: GET https://hey.mrwhale.ai/api/agent/v1/pricing
- OpenAPI: GET https://hey.mrwhale.ai/api/agent/openapi.json
- Brand discovery: https://mrwhale.ai/llms.txt
- Brand alias (same API): POST https://mrwhale.ai/api/agent/v1/images/generations

### How to call (Images API–style)

```json
{
  "prompt": "a small cute red whale icon, flat vector, white background",
  "size": "1024x1024"
}
```

| Field | Required | Default | Notes |
|-------|----------|---------|-------|
| prompt | yes | — | Text prompt; EN/ZH; same style as GPT Image 2 |
| size | no | 1024x1024 | One of the 1K sizes below |

#### Accepted sizes (1K MVP)
- `1024x1024` — 1:1 square
- `1024x1536` — 3:4 portrait
- `1536x1024` — 4:3 landscape
- `1024x1792` — 9:16 tall / stories
- `1792x1024` — 16:9 wide

#### Not supported yet
- n > 1
- image edits / image[] reference uploads
- 2K / 4K sizes
- custom quality (forced auto)
- response_format override (server returns hosted URLs)

#### curl (probe → expect 402)
```bash
curl -i -X POST https://hey.mrwhale.ai/api/agent/v1/images/generations \
  -H 'content-type: application/json' -H 'accept: application/json' \
  -d '{"prompt":"a red whale logo, flat vector","size":"1024x1024"}'
```

#### Paid flow
1. POST generate without payment → **402** + `PAYMENT-REQUIRED` header
2. Sign USDC with an x402 client (exact scheme on base)
3. Retry **same** POST + body with header `PAYMENT-SIGNATURE` → **202** `{ "jobId": "...", "poll": "..." }`
4. Poll `GET https://hey.mrwhale.ai/api/agent/v1/images/jobs/{jobId}` every 2–5s
5. When `status` is `done`, read image URL from `images[0]` or `result.data[0].url`

#### Job response (done)
```json
{
  "ok": true,
  "status": "done",
  "id": "gen-xxxx",
  "images": ["https://img.hey.mrwhale.ai/.../image.png"],
  "result": { "data": [{ "url": "https://img.hey.mrwhale.ai/.../image.png" }] }
}
```

Statuses: `pending` | `running` | `done` | `error` | `cancelled`

### Auth
- Agents: no API key. Pay the 402 challenge.
- Settlement may appear in `payment-response` response header after pay.

### Compatibility
Treat this like OpenAI **GPT Image 2** text-to-image for prompts and 1K sizes. Do not send `model` in the body — the server always uses GPT Image 2 / `jingtu-image-1k`.