Video Generation API

Video generation uses an asynchronous task model: create a task → poll the status → read the result. Video and 3D share the same task API and are routed by the model field.

There are currently no publicly discoverable video model call names in the model list or the Model Plaza. The video requests below only illustrate the structure and cannot be executed directly; <MODEL_CALL_NAME> must be replaced with a real call name published later on the Model Plaza.

Video generation uses the platform's own API + API key, not Volcengine AK/SK signing. The Volcengine-compatible APIs are for asset management only.

Key Parameters

Creating tasks: Authorization: Bearer <API_KEY> or x-api-key: <API_KEY>

Querying and canceling tasks: Authorization: Bearer <API_KEY>

https://www.silvamux.com/api/v3 is the /api/v3 path under the access domain, routing to /api/v3/contents/generations/tasks.

Creating a Task

POST https://www.silvamux.com/api/v3/contents/generations/tasks
Authorization: Bearer $SILVAMUX_API_KEY
Content-Type: application/json

Text-to-Video

curl -X POST https://www.silvamux.com/api/v3/contents/generations/tasks \
  -H "Authorization: Bearer $SILVAMUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<MODEL_CALL_NAME>",
    "content": [{"type": "text", "text": "A golden retriever running on the beach"}],
    "resolution": "720p",
    "ratio": "16:9",
    "duration": 5,
    "generate_audio": true
  }'

# pip install requests
import os
import requests

resp = requests.post(
    "https://www.silvamux.com/api/v3/contents/generations/tasks",
    headers={
        "Authorization": f"Bearer {os.environ['SILVAMUX_API_KEY']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "<MODEL_CALL_NAME>",
        "content": [{"type": "text", "text": "A golden retriever running on the beach"}],
        "resolution": "720p",
        "ratio": "16:9",
        "duration": 5,
        "generate_audio": True,
    },
)
resp.raise_for_status()
print(resp.json())  # {"id": "VTK-...", "status": "queued", ...}

// Node 18+ (built-in fetch)
const resp = await fetch(
  "https://www.silvamux.com/api/v3/contents/generations/tasks",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.SILVAMUX_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      model: "<MODEL_CALL_NAME>",
      content: [{ type: "text", text: "A golden retriever running on the beach" }],
      resolution: "720p",
      ratio: "16:9",
      duration: 5,
      generate_audio: true,
    }),
  },
);

const task = await resp.json();
console.log(task); // { id: "VTK-...", status: "queued", ... }

// Standard library only
package main

import (
    "bytes"
    "fmt"
    "io"
    "net/http"
    "os"
)

func main() {
    payload := []byte(`{
        "model": "<MODEL_CALL_NAME>",
        "content": [{"type": "text", "text": "A golden retriever running on the beach"}],
        "resolution": "720p",
        "ratio": "16:9",
        "duration": 5,
        "generate_audio": true
    }`)

    req, err := http.NewRequest("POST", "https://www.silvamux.com/api/v3/contents/generations/tasks", bytes.NewReader(payload))
    if err != nil {
        panic(err)
    }
    req.Header.Set("Authorization", "Bearer "+os.Getenv("SILVAMUX_API_KEY"))
    req.Header.Set("Content-Type", "application/json")

    resp, err := http.DefaultClient.Do(req)
    if err != nil {
        panic(err)
    }
    defer resp.Body.Close()

    body, _ := io.ReadAll(resp.Body)
    fmt.Println(string(body))
}

Image-to-Video

Image-to-video has two mutually exclusive scenarios: first frame, and first + last frames.

First Frame

curl -X POST https://www.silvamux.com/api/v3/contents/generations/tasks \
  -H "Authorization: Bearer $SILVAMUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<MODEL_CALL_NAME>",
    "content": [
      {"type": "image_url", "image_url": {"url": "https://example.com/first.jpg"}, "role": "first_frame"},
      {"type": "text", "text": "The camera slowly pulls back"}
    ]
  }'

First and Last Frames

curl -X POST https://www.silvamux.com/api/v3/contents/generations/tasks \
  -H "Authorization: Bearer $SILVAMUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<MODEL_CALL_NAME>",
    "content": [
      {"type": "image_url", "image_url": {"url": "https://example.com/first.jpg"}, "role": "first_frame"},
      {"type": "image_url", "image_url": {"url": "https://example.com/last.jpg"}, "role": "last_frame"},
      {"type": "text", "text": "The scene transitions from day to night"}
    ]
  }'

Multimodal-Reference Video

Generate video from reference images (1–9) + reference videos (0–3) + reference audio (0–3) + an optional text prompt, supporting fresh generation, editing, and extension.

curl -X POST https://www.silvamux.com/api/v3/contents/generations/tasks \
  -H "Authorization: Bearer $SILVAMUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<MODEL_CALL_NAME>",
    "content": [
      {"type": "image_url", "image_url": {"url": "asset://AST-XXXX"}, "role": "reference_image"},
      {"type": "text", "text": "Make the character in the frame dance"}
    ]
  }'

Audio cannot be the only input — include at least one reference video or image. The three image-to-video scenarios (first frame / first + last frames / multimodal reference) are mutually exclusive.

The content Field

Text (text)

Image (image_url)

Image requirements: formats jpeg/png/webp/bmp/tiff/gif; aspect ratio (0.4, 2.5); dimensions 300–6000px; each ≤30MB.

Video (video_url)

Video requirements: formats mp4/mov; resolutions 480p/720p/1080p; duration ≤15s, at most 3 reference videos totaling ≤15s; each ≤50MB; frame rate 4–60fps.

Audio (audio_url)

Audio cannot be the only input.

Audio requirements: formats wav/mp3; duration ≤15s, at most 3 clips totaling ≤15s; each ≤15MB.

Response

Creating a task returns a task ID:

{
  "id": "VTK-01JXXXXXXXXXXXXXX",
  "model": "<MODEL_CALL_NAME>",
  "status": "queued",
  "created_at": "2026-03-31T12:00:00Z"
}

Querying a Task

GET /api/v3/contents/generations/tasks/:id

GET https://www.silvamux.com/api/v3/contents/generations/tasks/{id}
Authorization: Bearer $SILVAMUX_API_KEY

curl https://www.silvamux.com/api/v3/contents/generations/tasks/VTK-01JXXXXXXXXXXXXXX \
  -H "Authorization: Bearer $SILVAMUX_API_KEY"

# pip install requests
import os
import requests

task_id = "VTK-01JXXXXXXXXXXXXXX"

resp = requests.get(
    f"https://www.silvamux.com/api/v3/contents/generations/tasks/{task_id}",
    headers={"Authorization": f"Bearer {os.environ['SILVAMUX_API_KEY']}"},
)
resp.raise_for_status()
print(resp.json())  # {"id": "...", "status": "succeed", "result": {...}}

// Node 18+ (built-in fetch)
const taskId = "VTK-01JXXXXXXXXXXXXXX";

const resp = await fetch(
  `https://www.silvamux.com/api/v3/contents/generations/tasks/${taskId}`,
  {
    headers: { Authorization: `Bearer ${process.env.SILVAMUX_API_KEY}` },
  },
);

const task = await resp.json();
console.log(task); // { id: "...", status: "succeed", result: {...} }

// Standard library only
package main

import (
    "fmt"
    "io"
    "net/http"
    "os"
)

func main() {
    taskID := "VTK-01JXXXXXXXXXXXXXX"

    req, err := http.NewRequest("GET", "https://www.silvamux.com/api/v3/contents/generations/tasks/"+taskID, nil)
    if err != nil {
        panic(err)
    }
    req.Header.Set("Authorization", "Bearer "+os.Getenv("SILVAMUX_API_KEY"))

    resp, err := http.DefaultClient.Do(req)
    if err != nil {
        panic(err)
    }
    defer resp.Body.Close()

    body, _ := io.ReadAll(resp.Body)
    fmt.Println(string(body))
}

Successful response:

{
  "id": "VTK-01JXXXXXXXXXXXXXX",
  "status": "succeed",
  "result": {"video_url": "https://..."}
}

Task statuses: queued → running → succeed / failed / cancelled / expired.

Task List

GET /api/v3/contents/generations/tasks

GET https://www.silvamux.com/api/v3/contents/generations/tasks
Authorization: Bearer $SILVAMUX_API_KEY

Canceling a Task

DELETE /api/v3/contents/generations/tasks/:id

DELETE https://www.silvamux.com/api/v3/contents/generations/tasks/{id}
Authorization: Bearer $SILVAMUX_API_KEY

Only queued tasks can be canceled.

Generation Showcase

Multimodal-reference video (background image + costume three-view sheet + face close-up + prompt):

Prompt:

Background reference image 1. A pale flash crosses the frame; the young nobleman (costume reference image 2; likeness strictly follows image 3) spins and snaps open a folding fan, gilded blades springing out as the ink-bamboo fan face flutters, a slow drum booming once. Close-up: the fan blade parries the villain's long saber, steel on steel; the nobleman's lips curl into a sly smile while his eyes stay cold, fingertips idly twirling the fan handle. Slow motion: he leans low and slides sideways, the fan blade grazing the villain's leg and leaving a faint mark, his robe hem sweeping the ground, jade hairpin swaying. Quick cut: he whirls and throws the fan — the blade shoots past the villain's neck and buries itself in the wooden pillar behind; the villain freezes, not daring to move. Reversal: a palm wind rushes in from behind; he spins to catch it, and as fingertips touch he borrows the force to leap back while the fan blade flies back to his hand from the pillar, eyes alert. Slow-motion highlight: fan half closed, blade resting by his lips, he glances back over his shoulder, hair lifted by the wind, brow raised with a trace of defiance. Pull-back: the nobleman stands on the stone terrace of a courtyard, lightly waving the fan as the camera pulls away, masked figures with curved sabers emerging from all four corners, closing in. Freeze frame: the fan half folded, gilded edge gleaming, he steps forward as the frame darkens, leaving only his silhouette and the blade's glint, sound cut to silence. SFX: crisp fan snaps + blade-wind hiss + slow drum hits (steady).

Available Models

The 3D sample model currently discoverable in the Model Plaza: doubao-seed3d-2.0. There are no publicly discoverable call names for video models yet.

The full model catalog is on the Model Plaza. Some models require permission flags configured by administrators.

Billing

Video generation is billed by model and video parameters (resolution, duration, audio); check current unit prices on the corresponding model detail page in the Model Plaza.

Differences from the Volcengine Official API

SilvaMux's Seedance API resembles the Volcengine official API field-wise (same models), but:

• Authentication: the platform API key (sk_live_), not Volcengine AK/SK.
• Base URL: https://www.silvamux.com/api/v3 (the platform domain), not the Volcengine Ark domain.
• model: the real call name shown by the Model Plaza or the model list, not a Volcengine endpoint ID.
• Task queries: poll GET /api/v3/contents/generations/tasks/:id for task status.