# Cosmos 3 Super Image to Video API

> Anima uma imagem do primeiro quadro em um vídeo, com expansão opcional do prompt e um ciclo agentivo de refinamento

Usa `image_url` como quadro inicial e um `prompt` de até 4096 caracteres. `num_frames` chega a 189 e `frames_per_second` varia de 4 a 60. A expansão do prompt vem ativada por padrão, e `enable_agentic_generation` executa candidatos iterativos com crítica. Também oferece `negative_prompt`, `guidance_scale` e `num_inference_steps` de até 50.

- Model id: `nvidia/cosmos-3-super/image-to-video`
- Made by: NVIDIA
- Category: Image to Video
- Output: video
- Price per second of your output video: from $0.05 / second of output video (cheapest run $0.05 (50 API credits))
- Run: `POST https://api.doitong.ru/v1/run/nvidia/cosmos-3-super/image-to-video`
- Quote: `POST https://api.doitong.ru/v1/quote/nvidia/cosmos-3-super/image-to-video`
- Docs page: https://doitong.com/pt-BR/developers/models/nvidia/cosmos-3-super/image-to-video

## Pricing

Price: **from $0.05 / second of output video** (the lowest tier; the quote prices your exact input). Price per second of your output video: the input files are measured before the run (length, resolution, frame rate). The quote measures the same files and is exactly what the run charges.

The full example input below costs $0.40 = 400 API credits (8 s output video × $0.05).

Prices are the upstream price in USD, paid from the API balance in API credits (1000 API credits = 1 USD). `POST https://api.doitong.ru/v1/quote/nvidia/cosmos-3-super/image-to-video` with the same JSON body returns the exact price for that input; the quote is exactly what a run charges. Invalid input is never charged and failed or canceled requests are refunded automatically.

## Input parameters

Send either the canonical name or the native field name (not both). Canonical names are the same across models.

| Parameter | Canonical name | Type | Required | Default | Allowed values / range | Description |
|---|---|---|---|---|---|---|
| `prompt` |  | string | yes |  | 1 – 4096 characters | Instruções de geração de vídeo em texto: especifique o que a cena contém e o movimento que ela deve mostrar. |
| `image_url` | `image` | string (image URL) | yes |  |  | URL da imagem usada para condicionar o primeiro quadro do vídeo. |
| `negative_prompt` |  | string | no | `The video captures a series of frames showing macroblocking artifacts, chromatic aberration, high-frequency noise, and rolling shutter distortion. It includes static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, bit-depth compression artifacts, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, hard cut, visual noise, and flickering. It features moiré patterns, edge halos, and temporal aliasing. Furthermore, the content defies common sense, generating illogical scenarios, nonsensical entities, absurd character behaviors, and conceptual paradoxes that violate basic human reasoning and everyday reality. The video looks like a surreal or glitchy hallucination. Overall, the video is of poor quality.` | ≤ 2048 characters | Prompt negativo para desencorajar conteúdo como movimentos ou artefatos indesejados. Omiti-lo usa o prompt negativo recomendado pela NVIDIA para i2v, enquanto uma string vazia desativa a orientação negativa. |
| `enable_prompt_expansion` |  | boolean | no | `true` |  | Expansão de prompt que, quando true, usa Cosmos3-Nano Reasoner para produzir o formato de legenda densa usado no treinamento do Cosmos3; esse VLM examina o primeiro quadro. COSMOS_PROMPT_UPSAMPLER_BASE_URL seleciona um serviço Reasoner configurado; caso contrário, o aplicativo inicia um localmente. Se a expansão falhar, o prompt sem modificações continuará sendo usado. |
| `enable_agentic_generation` |  | boolean | no | `false` |  | Geração agentiva iterativa do Cosmos, consideravelmente mais lenta e cara que gerar uma vez, pois cada candidato exige uma renderização completa. O loop amplia o prompt, gera candidatos de vídeo, usa um VLM para criticar quadros amostrados e reescreve o prompt. |
| `agentic_max_iterations` |  | integer | no | `2` | 1 – 3 | Limite superior de etapas do prompt com a geração agentiva ativada. |
| `agentic_samples_per_iteration` |  | integer | no | `2` | 1 – 3 | Quantidade de candidatos de vídeo por iteração no loop agentivo, determinando quantos vídeos competem na avaliação antes que o melhor avance para a reescrita do prompt. |
| `agentic_early_stop` |  | boolean | no | `true` |  | Controle de parada antecipada que considera uma pontuação do avaliador acima do limite rigoroso de qualidade suficiente para encerrar o loop agentivo. |
| `image_size` |  | object {width, height} \| enum (string) | no | `{"height":480,"width":832}` | `square_hd`, `square`, `portrait_4_3`, `portrait_16_9`, `landscape_4_3`, `landscape_16_9` | Dimensões do vídeo solicitadas antes do ajuste automático. Para obter um tamanho compatível, o sistema limita a requisição e escolhe o nível NVIDIA mais próximo — 256p, 480p ou 720p — e uma proporção compatível. |
| `num_frames` |  | integer | no | `189` | 5 – 189 | Quantidade de quadros para a geração; aumentá-la produz um vídeo mais longo. |
| `frames_per_second` |  | integer | no | `24` | 4 – 60 | Taxa de quadros do vídeo de saída, medida em quadros por segundo. |
| `num_inference_steps` |  | integer | no | `28` | 1 – 50 | Quantidade de etapas de remoção de ruído, trocando processamento mais longo por melhor qualidade à medida que a quantidade aumenta. |
| `guidance_scale` |  | number | no | `6` | 0 – 20 | Intensidade da orientação sem classificador. |
| `seed` |  | integer \| null | no |  |  | Seed que controla a repetibilidade: um prompt e uma seed idênticos produzem um vídeo idêntico quando a versão do modelo não muda. |
| `enable_safety_checker` |  | boolean | no | `true` |  | Controle de moderação para a imagem e o prompt fornecidos, cuja desativação é permitida somente para contas autorizadas. Requisições sem autorização passam por verificações independentemente da configuração. |

File parameters (`image`) take a public `https://` URL, a `data:` URI up to 10 MB, or the `file_url` of an upload from `POST /files`.

## Example input

Required parameters only:

```json
{
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
  "image": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
}
```

A fuller example:

```json
{
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
  "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
}
```

## Run it

Set `DOITONG_API_KEY` to a key from https://doitong.com/developers/dashboard#keys. `?wait=20` returns the finished request in the same call when it is ready within 20 seconds; otherwise poll `GET /requests/{id}` (it also accepts `?wait=`), or pass `webhook_url`.

### curl

```bash
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.ru/v1/run/nvidia/cosmos-3-super/image-to-video?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
  }'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.ru/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```

### JavaScript

```javascript
const API = "https://api.doitong.ru/v1";
const headers = {
  "Authorization": "Bearer " + process.env.DOITONG_API_KEY,
  "Content-Type": "application/json",
};

// Run the model; wait up to 20 s for the finished result in the same call
let res = await fetch(API + "/run/nvidia/cosmos-3-super/image-to-video?wait=20", {
  method: "POST",
  headers,
  body: JSON.stringify({
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
  }),
});
let request = await res.json();
if (!res.ok) throw new Error(request.error.code + ": " + request.error.message);

// Not finished yet: long-poll until it is
while (request.status === "queued" || request.status === "running") {
  res = await fetch(API + "/requests/" + request.id + "?wait=20", { headers });
  request = await res.json();
}
if (request.status !== "succeeded") throw new Error(JSON.stringify(request.error));
console.log(request.outputs.map((o) => o.url), request.cost.credits + " API credits ($" + request.cost.usd + ")");
```

### Python

```python
import os
import requests

API = "https://api.doitong.ru/v1"
headers = {"Authorization": "Bearer " + os.environ["DOITONG_API_KEY"]}

payload = {
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg",
}

# Run the model; wait up to 20 s for the finished result in the same call
r = requests.post(API + "/run/nvidia/cosmos-3-super/image-to-video", params={"wait": 20}, json=payload, headers=headers)
request = r.json()
if not r.ok:
    raise RuntimeError(request["error"]["code"] + ": " + request["error"]["message"])

# Not finished yet: long-poll until it is
while request["status"] in ("queued", "running"):
    request = requests.get(API + "/requests/" + request["id"], params={"wait": 20}, headers=headers).json()
if request["status"] != "succeeded":
    raise RuntimeError(request["error"])
print([o["url"] for o in request["outputs"]], request["cost"]["credits"], "API credits", "$" + request["cost"]["usd"])
```

### MCP

The MCP server at `https://api.doitong.ru/v1/mcp` exposes the tools `search_models`, `get_model`, `quote`, `run`, `get_request`, `cancel_request`, `list_requests`, `create_upload` and `get_account`. A `run` call for this model:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "run",
    "arguments": {
      "model_id": "nvidia/cosmos-3-super/image-to-video",
      "input": {
        "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
        "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
      },
      "wait_seconds": 20
    }
  }
}
```

## Response

Every call returns a request object. `status` is one of `queued`, `running`, `succeeded`, `failed`, `canceled`. `outputs` lists the generated files with URLs on our storage (kept forever, or 7 days with `?storage=temporary`; `storage` says which); `output` is the model's own result object with the same URLs.

```json
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "nvidia/cosmos-3-super/image-to-video",
  "status": "succeeded",
  "created_at": "2026-10-20T12:00:00Z",
  "started_at": "2026-10-20T12:00:02Z",
  "completed_at": "2026-10-20T12:00:09Z",
  "queue_position": null,
  "cost": {
    "credits": 400,
    "usd": "0.40",
    "charged": 400,
    "refunded": 0
  },
  "input": {
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
    "image_url": "https://content.doitong.com/images/showcase/v3/gen_fox_cartoon.jpg"
  },
  "output": {
    "video": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_name": "0-output.mp4",
      "file_size": 1843200,
      "width": 1024,
      "height": 768,
      "fps": 1.0,
      "duration": 5,
      "num_frames": 1
    },
    "seed": 42
  },
  "outputs": [
    {
      "type": "video",
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.mp4",
      "content_type": "video/mp4",
      "file_size": 1843200,
      "width": 1280,
      "height": 720,
      "duration": 5,
      "stored": true
    }
  ],
  "error": null,
  "metrics": {
    "inference_seconds": 6.8
  },
  "urls": {
    "self": "https://api.doitong.ru/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t",
    "cancel": "https://api.doitong.ru/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel"
  }
}
```

### Output fields (`output`)

| Field | Type | Description |
|---|---|---|
| `video` | object {url, content_type, file_name, file_size, width, height, fps, duration, num_frames} | Vídeo produzido pelo modelo. |
| `seed` | integer | Valor de seed aplicado durante a geração. |

## Related models

- [Cosmos 3 Super](https://doitong.com/developers/models/nvidia/cosmos-3-super/text-to-image.md) (`nvidia/cosmos-3-super/text-to-image`): from $0.04 / run of output image
- [FLUX 3 Max Image to Video](https://doitong.com/developers/models/blackforestlabs/flux-3-max/image-to-video.md) (`blackforestlabs/flux-3-max/image-to-video`): from $0.0425 / second of output video
- [Grok Imagine Video 1.5 Image to Video](https://doitong.com/developers/models/xai/grok-imagine-video/v1.5/image-to-video.md) (`xai/grok-imagine-video/v1.5/image-to-video`): from $0.082 / second
- [Kling 3.0 Pro Image to Video](https://doitong.com/developers/models/kling-video/v3/pro/image-to-video.md) (`kling-video/v3/pro/image-to-video`): from $0.112 / second
- [LTX 2.5 Pro Image to Video](https://doitong.com/developers/models/lightricks/ltx-2.5/image-to-video/pro.md) (`lightricks/ltx-2.5/image-to-video/pro`): from $0.12 / second
- [MiniMax H3 Image to Video](https://doitong.com/developers/models/minimax/h3/image-to-video.md) (`minimax/h3/image-to-video`): from $0.05 / second
- [Seedance 2.0 Fast Image to Video](https://doitong.com/developers/models/bytedance/seedance-2.0/fast/image-to-video.md) (`bytedance/seedance-2.0/fast/image-to-video`): from $0.113 / second
- [Seedance 2.0 Image to Video](https://doitong.com/developers/models/bytedance/seedance-2.0/image-to-video.md) (`bytedance/seedance-2.0/image-to-video`): from $0.141 / second

## Links

- HTML page: https://doitong.com/developers/models/nvidia/cosmos-3-super/image-to-video
- All models and guides for AI agents: https://doitong.com/developers/llms.txt
- OpenAPI 3.1 with this model's typed request body: https://doitong.com/developers/openapi.json?model=nvidia/cosmos-3-super/image-to-video
- Errors: https://doitong.com/developers/docs/errors
- Model details as JSON: https://api.doitong.ru/v1/models/nvidia/cosmos-3-super/image-to-video