What Is Text-to-Video?

Seedance 2.0's text-to-video capability turns natural language descriptions into high-quality cinematic video. Describe the scene, motion, and style you want, and the model generates a matching video.

Model ID: seedance-2.0-text-to-video (for faster generation and lower cost, use seedance-2.0-fast-text-to-video)

How It Works

Write the prompt — describe the scene you want
Send the API request — model, prompt, quality, duration, and other parameters
Get a task ID — poll asynchronously or configure a callback_url webhook
Download the video — retrieve the URL when generation completes (valid 24 hours)

Prompt Engineering Tips

Be Specific

Don't just say "a dog running". Write: "A golden retriever running through a sunlit meadow, slow motion, shallow depth of field, cinematic."

Use Style Keywords

Shots: aerial shot, close-up, tracking shot, dolly zoom
Lighting: golden hour, studio lighting, neon, dramatic shadows
Style: cinematic, slow motion, timelapse, documentary

Structure Your Prompt

[Subject] + [Action] + [Setting] + [Style/Mood]

Example: A woman in a red dress walking through a rain-soaked Tokyo street at night, neon reflections on wet pavement, cinematic wide shot

Put Dialogue in Double Quotes

Wrap spoken lines in straight double quotes. The model detects them as dialogue and optimizes speech synthesis: She turned and said: "You finally understood."

API Reference

import requests

response = requests.post(
    "https://api.evolink.ai/v1/videos/generations",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "seedance-2.0-text-to-video",
        "prompt": "Your scene description here",
        "quality": "720p",       # 480p or 720p
        "duration": 5,           # 4–15 seconds (integer)
        "aspect_ratio": "16:9",
        "generate_audio": True
    }
)

Web Search (Optional)

model_params.web_search is exclusive to the text-to-video family. When enabled, the model autonomously decides whether to search the internet for fresh content and is only billed when a search is actually triggered:

json={
    "model": "seedance-2.0-text-to-video",
    "prompt": "An ad for the latest 2026 electric sports car",
    "model_params": {"web_search": True}
}

See the full parameter reference in the Text-to-Video API docs.

Use Cases

E-commerce: Generate product showcase videos from descriptions
Marketing: Create ad creatives and social media content at scale
Entertainment: Rapid prototyping of scenes and storyboards
Education: Illustrate abstract concepts with custom videos

Seedance 2.0 Text-to-Video — Generate AI Video from Text Prompts