Logo for AiToolGo

Seedance 2.0 API: A Complete Guide to AI Text-to-Video and Image-to-Video Generation

In-depth discussion
Technical
 0
 0
 1
This README documents the Seedance 2.0 API, ByteDance's video generation model accessed via fal.ai. It covers three generation modes—text-to-video, image-to-video, and reference-to-video (multi-modal with image, video, and audio references)—with standard and fast pricing tiers. The document provides parameter tables, Python/JavaScript/cURL quick-start code, character-consistency, style transfer, lip-sync, and motion-transfer examples, plus a prompt-writing guide with cinematic, product, nature, anime, and music-video samples.
  • main points
  • unique insights
  • practical applications
  • key topics
  • key insights
  • learning outcomes

  • • main points

    • 1
      Complete API reference with parameter tables, types, defaults, and pricing for all six endpoints
    • 2
      Runnable code in three languages (Python, JavaScript, cURL) plus linked full examples
    • 3
      Dedicated prompt guide with a clear formula (Subject + Action + Camera + Scene/Lighting + Style) and domain-specific sample prompts

  • • unique insights

    • 1
      Practical prompt tips such as avoiding the word 'fast' to reduce jitter and describing camera and subject movement separately to prevent shaky output
    • 2
      The @Image1/@Video1/@Audio1 reference syntax combined with scene-cut and timestamp directives to control multi-scene pacing

  • • practical applications

    • Developers integrating AI video generation can move from zero to a working API call quickly, and the prompt guide and multi-modal examples help them achieve consistent characters, style transfer, and lip-synced speech without trial-and-error.

  • • key topics

    • 1
      Text-to-video, image-to-video, and reference-to-video generation endpoints and parameters
    • 2
      Multi-modal reference inputs (images, videos, audio) for character consistency, style transfer, lip-sync, and motion transfer
    • 3
      Prompt engineering for cinematic camera control, lighting, and multi-scene composition

  • • key insights

    • 1
      Unified omni-modal endpoint combining image, video, and audio references in a single generation call
    • 2
      Standard vs. fast tiers with transparent per-second pricing, enabling cost-quality tradeoffs
    • 3
      Prompt guide with concrete, copy-ready examples across multiple genres

  • • learning outcomes

    • 1
      Make successful authenticated API calls to generate text-to-video and image-to-video clips in Python, JavaScript, or cURL
    • 2
      Use @Image, @Video, and @Audio references with scene cuts and timestamps to create character-consistent, style-transferred, or lip-synced videos
    • 3
      Write structured, cinematic prompts that specify camera movement, lighting, and style to improve generation quality
examples
tutorials
code samples
visuals
fundamentals
advanced content
practical tips
best practices

“ Introduction to Seedance 2.0

Seedance 2.0 is ByteDance's advanced video generation model, now available through fal.ai as a robust text to video API and image to video API. It enables developers and creators to produce cinematic-quality clips from text prompts, static images, or multi-modal references. With native audio synthesis, real-world physics simulation, and precise camera control, Seedance 2.0 is designed for professional-grade AI video generation.

“ Core Features and Capabilities

Seedance 2.0 offers text-to-video generation for cinematic output, image-to-video animation that brings static images to life with realistic motion, and reference-to-video generation combining images, videos, and audio. Key capabilities include synchronized sound effects and lip-synced speech, director-level camera control, clip durations from 4 to 15 seconds at 480p or 720p, and support for aspect ratios including 16:9, 9:16, 4:3, 3:4, 1:1, and 21:9 ultrawide.

“ Available Endpoints and Pricing

The API is offered in two tiers. The standard tier delivers higher quality: text-to-video and image-to-video at $0.3034 and $0.3024 per second respectively, with reference-to-video at $0.3024 per second. The fast tier reduces cost and latency, priced at $0.2419 per second across text-to-video, image-to-video, and reference-to-video. Reference-to-video requests that include video inputs receive a 0.6x price multiplier.

“ Getting Started: Quick Start Integration

Integration is straightforward. In Python, install the client with `pip install fal-client` and call `fal_client.subscribe` with the model ID `bytedance/seedance-2.0/text-to-video`, passing a prompt, resolution, duration, aspect ratio, and audio flag. In JavaScript, install `@fal-ai/client` and use `fal.subscribe`. Developers can also call the endpoint directly via cURL with an Authorization key header, receiving a JSON response containing the generated video URL and seed.

“ Mastering the API Parameters

Core parameters include `prompt` (required), `resolution` (480p or 720p), `duration` ("auto" or 4-15 seconds), `aspect_ratio`, `generate_audio` (default true), `seed` for reproducibility, and `end_user_id` for tracking. The image-to-video endpoint adds `image_url` and an optional `end_image_url` for start-to-end frame transitions. Images must be JPEG, PNG, or WebP up to 30 MB.

“ Advanced Reference-to-Video Techniques

The reference-to-video endpoint is the most flexible option, accepting up to 9 reference images, 3 reference videos (2-15 seconds combined, up to 50 MB each), and 3 audio files (15 seconds combined). Using @Image1, @Video1, and @Audio1 syntax in prompts, developers can achieve character consistency, style transfer, lip-sync speech, motion transfer from video references, and multi-scene composition with scene cuts. The total number of files across all modalities must not exceed 12.

“ Prompt Writing Best Practices

Effective prompts follow the structure Subject + Action + Camera + Scene/Lighting + Style, typically 50-150 words. Be specific about camera movement ("slow dolly push in" rather than "camera moves forward") and always describe lighting, which has the largest impact on quality. Use "Cut scene to..." for multi-scene videos, avoid the word "fast" to reduce jitter, and describe camera and subject movement separately to prevent shaky output.

“ Practical Use Cases and Prompt Examples

Seedance 2.0 suits a range of applications: cinematic trailers and film noir scenes, product showcases and food commercials, talking-head videos with lip-synced audio, dance and music videos using motion transfer, nature timelapses, and anime-style stylized content. Each use case benefits from tailored prompts that specify camera behavior, lighting, color grading, and framing to achieve the intended cinematic or commercial look.

“ Authentication and Output Handling

Authenticate by exporting your fal API key as the `FAL_KEY` environment variable; the official client libraries read it automatically. All endpoints return a consistent response containing a `video.url` pointing to the generated MP4 file and the `seed` used, which can be reused to reproduce results.

“ Resources and Next Steps

To get started, try Seedance 2.0 in the fal.ai playground, consult the fal.ai documentation and Python and JavaScript client references, and review full pricing details. Complete runnable examples for text-to-video, image-to-video, and reference-to-video are available in both Python and JavaScript in the official repository, providing a practical foundation for building AI video generation into your applications.

 Original link: https://github.com/fal-ai/seedance-2.0-api

Comment(0)

user's avatar

      Similar Learning

      Related Tools