bytedance/seedance-2.0/reference-to-video

Run Seedance 2.0 AI Reference To Video API on fal

ByteDance's most advanced reference-to-video model, available on fal as `bytedance/seedance-2.0/reference-to-video`.

Overview

The most flexible Seedance 2.0 endpoint: provide a text prompt alongside up to 12 reference files spanning images, videos, and audio clips, and the model weaves them into a single cinematic output. Reference assets in your prompt using `@Image1`, `@Video1`, `@Audio1`, etc.

Key capabilities:

• Native audio generation: music, SFX, and lip-synced dialogue, all in a single pass at no extra cost
• Director-level camera control via prompt: dolly zooms, tracking shots, POV switches, handheld movement
• Realistic physics: fluid dynamics, cloth, character motion
• Multi-shot editing: natural cuts within a single generation, up to 15 seconds
• Output up to 1080p

Inputs

Total files across all modalities must not exceed 12.

Parameters

Pricing

Billed per second of generated 720p output:

Token formula (note: includes both input and output duration):

tokens = (height × width × (input_duration + output_duration) × 24) / 1024

Audio generation is included at no extra cost regardless of the `generate_audio` setting.

Quick Start

Python

bash
pip install fal-client
export FAL_KEY="YOUR_API_KEY"

python
import fal_client

result = fal_client.subscribe(
    "bytedance/seedance-2.0/reference-to-video",
    arguments={
        "prompt": "@Image1 shows a product on a marble surface. Slow dolly in with dramatic side lighting, dust particles floating in the air.",
        "image_urls": ["https://your-host.com/product.jpg"],
        "resolution": "1080p",
        "duration": "8",
        "aspect_ratio": "16:9",
        "generate_audio": True,
    },
    with_logs=True,
    on_queue_update=lambda u: [print(l["message"]) for l in u.logs]
    if isinstance(u, fal_client.InProgress) else None,
)

print(result["video"]["url"])

Multi-modal example with images, video, and audio:

python
result = fal_client.subscribe(
    "bytedance/seedance-2.0/reference-to-video",
    arguments={
        "prompt": "Recreate the scene from @Video1 but replace the background with the environment from @Image1. Use @Audio1 as the soundtrack.",
        "image_urls": ["https://your-host.com/background.jpg"],
        "video_urls": ["https://your-host.com/scene.mp4"],
        "audio_urls": ["https://your-host.com/soundtrack.mp3"],
        "resolution": "720p",
        "duration": "auto",
        "aspect_ratio": "auto",
    },
)

JavaScript / Node.js

bash
npm install @fal-ai/client
export FAL_KEY="YOUR_API_KEY"

js
import { fal } from "@fal-ai/client";

const result = await fal.subscribe("bytedance/seedance-2.0/reference-to-video", {
  input: {
    prompt: "@Image1 shows a product on a marble surface. Slow dolly in with dramatic side lighting, dust particles floating in the air.",
    image_urls: ["https://your-host.com/product.jpg"],
    resolution: "1080p",
    duration: "8",
    aspect_ratio: "16:9",
    generate_audio: true,
  },
  logs: true,
  onQueueUpdate: (update) => {
    if (update.status === "IN_PROGRESS") {
      update.logs.map((log) => log.message).forEach(console.log);
    }
  },
});

console.log(result.data.video.url);

Output

json
{
  "video": {
    "url": "https://...",
    "content_type": "video/mp4",
    "file_name": "output.mp4",
    "file_size": 4823041
  },
  "seed": 42
}

Async / Queue Usage

python
handler = fal_client.submit(
    "bytedance/seedance-2.0/reference-to-video",
    arguments={...},
    webhook_url="https://your-server.com/webhook",
)

request_id = handler.request_id
status = fal_client.status("bytedance/seedance-2.0/reference-to-video", request_id, with_logs=True)
result = fal_client.result("bytedance/seedance-2.0/reference-to-video", request_id)

Standard vs. Fast Tier

The only functional difference between the two tiers is resolution support and cost. Use fast unless you need 1080p.

Enterprise Variant

A separate enterprise endpoint adds face-input support for consistent character identity across generations:

fal-ai/seedance-2/enterprise/reference-to-video

Availability

• April 2, 2026: Launched with enterprise-only, geo-restricted access
• April 9, 2026: All restrictions lifted, fully open with no geographic or use-case limitations

Resources

• fal Model Page
• Try in Sandbox
• fal File Upload Guide
• fal Server-Side Integration Guide