> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mountsea.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

# Introduction to Hub

**Hub** is Mountsea's premium **AI Gateway** — a single unified API for the world's best image, video, audio and transcription models.

<Tip>
  **One endpoint per capability, many models.** Switch models without changing your code — just change the `model` field in the request body.
</Tip>

## Why Hub?

<CardGroup cols={3}>
  <Card title="Flagship Quality" icon="award">
    **Official flagship models only.** No knock-offs, no lossy distillations. Same model weights, same outputs — Veo 3.1, Nano Banana Pro, GPT Image 2, Kling v3 Pro, WAN 2.7, Seedance 2.0, ElevenLabs Music.
  </Card>

  <Card title="Production Stable" icon="shield-check">
    Built for **24/7 production traffic**. Multi-region routing, automatic failover, transparent retries, and queue-aware load balancing. No surprise rate-limit walls, no flaky upstreams.
  </Card>

  <Card title="Cheaper Than Official" icon="circle-dollar-to-slot">
    **Pay only for successful generations**, billed in unified credits at a meaningful discount versus going direct to the model provider. No per-provider minimums, no monthly subscriptions.
  </Card>
</CardGroup>

### A Closer Look

|                 | Hub                                                                                             | Going direct to each provider                                                                         |
| --------------- | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
| **Models**      | Official flagship models across 4 capabilities — image, video, audio, transcribe                | Sign up & maintain 6+ separate accounts (Google, OpenAI, Kuaishou, Alibaba, ByteDance, ElevenLabs, …) |
| **Pricing**     | **Lower than the official list price**, paid only on `status=completed` — failed tasks are free | Pay full retail; failures still consume quota on most providers                                       |
| **Stability**   | Smart routing across redundant upstream channels, automatic retry on transient errors           | Single point of failure; manual retry logic & rate-limit handling on your side                        |
| **Onboarding**  | One Bearer token, one Base URL, one credit balance                                              | Per-provider keys, per-provider billing, per-provider SDKs                                            |
| **Maintenance** | New models added & old ones routed for you                                                      | Track every provider's deprecations & migration notices yourself                                      |

<Tip>
  **Same outputs at a lower price.** Hub forwards your request to the original provider's flagship endpoint — what comes back is the same media you'd get going direct, just at Hub's discounted rate and with stability guarantees baked in.
</Tip>

## Capabilities

<CardGroup cols={2}>
  <Card title="Image" icon="image" href="/api-reference/hub/image">
    **Nano Banana** (Fast / 2 / Pro) + **GPT Image 2** — text-to-image and image editing
  </Card>

  <Card title="Video" icon="video" href="/api-reference/hub/video">
    **Veo 3.1** · **Kling v3** · **WAN 2.7** · **Seedance 2.0** — t2v, i2v, multi-ref, first-last, video edit
  </Card>

  <Card title="Audio (Music)" icon="music" href="/api-reference/hub/audio">
    **ElevenLabs Music** — text-to-music with length & instrumental control
  </Card>

  <Card title="Transcribe" icon="microphone" href="/api-reference/hub/transcribe">
    **Whisper / Wizper** — audio & video speech-to-text and translation
  </Card>
</CardGroup>

## The Hub Pattern

Every Hub task follows the same simple pattern. Submit a task → get a `task_id` → poll until `ready=true`.

<Steps>
  <Step title="Discover available models (optional)">
    Call `GET /hub/v1/models?capability=image|video|audio|transcribe` to see all models for a capability.
    For a specific model, `GET /hub/v1/models/:model` returns its full `input_schema` plus a ready-to-copy `example` payload.
  </Step>

  <Step title="Submit a task">
    Send `POST /hub/v1/{image|video|audio|transcribe}` with `{ model, input }`. You get back `{ task_id }`.
  </Step>

  <Step title="Poll for the result">
    `GET /hub/v1/tasks/:task_id` — poll every 3–5 seconds until `ready=true`. The result is in `data`.
  </Step>
</Steps>

## Featured Models

### 🖼️ Image

| Model             | Provider | Capability     | Highlights                                     |
| ----------------- | -------- | -------------- | ---------------------------------------------- |
| `nano-banana`     | Google   | text-to-image  | Gemini 2.5 Flash Image — fast & cheap          |
| `nano-banana-2`   | Google   | text-to-image  | Gemini 3.1 — extreme aspect ratios (1:8 / 8:1) |
| `nano-banana-pro` | Google   | text-to-image  | Gemini 3 Pro — studio quality, up to 4K        |
| `gpt-image-2`     | OpenAI   | text-to-image  | Detailed images with fine typography           |
| `*-edit` variants | —        | image-to-image | Edit existing images with reference URLs       |

### 🎬 Video

| Model                         | Provider  | Capability          | Highlights                           |
| ----------------------------- | --------- | ------------------- | ------------------------------------ |
| `veo-3.1` / `-fast` / `-lite` | Google    | text-to-video       | Native audio, 4s / 6s / 8s, up to 4K |
| `veo-3.1-image`               | Google    | image-to-video      | Animate a single reference image     |
| `veo-3.1-ref`                 | Google    | reference-to-video  | Multi-image consistent character     |
| `veo-3.1-first-last`          | Google    | first-last frame    | Transition between two frames        |
| `kling-v3-pro` / `-standard`  | Kuaishou  | text/image-to-video | 3–15s, native audio, multi-shot      |
| `wan-2.7`                     | Alibaba   | text-to-video       | High quality, default 1080p          |
| `wan-2.7-image`               | Alibaba   | image-to-video      | First-and-last frame, audio driving  |
| `wan-2.7-ref`                 | Alibaba   | reference-to-video  | Multi-subject reference              |
| `wan-2.7-edit`                | Alibaba   | video-to-video      | Instruction-based video editing      |
| `seedance-2.0` / `-fast`      | ByteDance | text/image-to-video | Cinematic, native audio, physics     |

### 🎵 Audio (Music)

| Model              | Provider   | Capability     | Highlights                                   |
| ------------------ | ---------- | -------------- | -------------------------------------------- |
| `elevenlabs-music` | ElevenLabs | music-generate | Text-to-music, 3s–10min, instrumental switch |

### 🎙️ Transcribe

| Model            | Provider | Capability             | Highlights                                       |
| ---------------- | -------- | ---------------------- | ------------------------------------------------ |
| Whisper / Wizper | —        | transcribe / translate | BCP-47 language codes, word / segment timestamps |

> Use `GET /hub/v1/models?capability=transcribe` for the up-to-date list.

***

## Quick Example — Image

```bash theme={null}
# 1. Submit
curl -X POST "https://api.mountsea.ai/hub/v1/image" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-pro",
    "input": {
      "prompt": "A photorealistic black labrador swimming, half above water, half underwater",
      "resolution": "2K",
      "aspect_ratio": "1:1"
    }
  }'
# → { "task_id": "hub-xxxxxxxx-..." }

# 2. Poll
curl -H "Authorization: Bearer your-api-key" \
  "https://api.mountsea.ai/hub/v1/tasks/hub-xxxxxxxx-..."
```

## Quick Example — Video

```bash theme={null}
curl -X POST "https://api.mountsea.ai/hub/v1/video" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo-3.1-fast",
    "input": {
      "prompt": "Cinematic drone shot flying through ancient stone ruins at golden hour",
      "duration": "8s",
      "resolution": "720p",
      "aspect_ratio": "16:9",
      "generate_audio": true
    }
  }'
```

## Quick Example — Music

```bash theme={null}
curl -X POST "https://api.mountsea.ai/hub/v1/audio" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "elevenlabs-music",
    "input": {
      "prompt": "Uplifting lo-fi hip-hop with mellow piano and vinyl crackle",
      "music_length_ms": 30000
    }
  }'
```

## Quick Example — Transcribe

```bash theme={null}
curl -X POST "https://api.mountsea.ai/hub/v1/transcribe" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "whisper-v3",
    "input": {
      "audio_url": "https://example.com/audio.mp3",
      "language": "en",
      "timestamps": "word"
    }
  }'
```

***

## Endpoints at a Glance

| Endpoint                  | Method | Description                                          |
| ------------------------- | ------ | ---------------------------------------------------- |
| `/hub/v1/image`           | POST   | Submit an image generation / edit task               |
| `/hub/v1/video`           | POST   | Submit a video generation / edit task                |
| `/hub/v1/audio`           | POST   | Submit a music generation task                       |
| `/hub/v1/transcribe`      | POST   | Submit a transcription / translation task            |
| `/hub/v1/tasks/{task_id}` | GET    | Poll task status & result                            |
| `/hub/v1/models`          | GET    | List every model (optionally filtered by capability) |
| `/hub/v1/models/{model}`  | GET    | Get a single model's full input schema + example     |

## Task Status

| Status       | Meaning                                       |
| ------------ | --------------------------------------------- |
| `pending`    | Queued, waiting for a worker                  |
| `processing` | Actively running                              |
| `completed`  | ✅ Done — `data` contains the result           |
| `failed`     | ❌ Failed — see `error_code` / `error_message` |
| `timeout`    | Exceeded the processing time limit            |
| `cancelled`  | Cancelled by user or system                   |

<Tip>
  Recommended polling interval: **3–5 seconds**. Generated assets are stored for **two weeks** — download promptly.
</Tip>

## Authentication

All endpoints require Bearer token authentication:

```
Authorization: Bearer your-api-key
```

## Base URL

```
https://api.mountsea.ai
```

***

### Explore the API Documentation

* [Submit Image Task](image) — Nano Banana, GPT Image 2 (+ edit variants)
* [Submit Video Task](video) — Veo 3.1, Kling v3, WAN 2.7, Seedance 2.0
* [Submit Audio Task](audio) — ElevenLabs Music
* [Submit Transcribe Task](transcribe) — Whisper / Wizper
* [Poll Task Result](task) — Get task status & result
* [List Models](models) — Browse all available models
* [Get Model Details](model-detail) — Inspect input schema + example