LESSON 3 · PRACTICE

Your first generation: text → video

It's time to get your first result with your own hands. The simplest path in Seedance 2.0 is the "text→video" mode: you describe the scene in words, and the model turns it into a clip. No uploads, no references — just text. In this lesson you'll walk the whole way from an empty prompt field to a finished video and learn to improve it calmly.

🎯 What you'll do
Create a short 5-second video from a single paragraph of text, evaluate the result, and make your first edit. This is the basic cycle you'll repeat in every lesson that follows.

Why text→video is the right start

Text is the model's native language. It works best for spatial decisions: what the object looks like, what the mood is, what the frame's style is. In this mode the model isn't tied to uploaded images or videos, so you'll quickly grasp how your words shape the picture. References (photo, video, audio) come into play later, when you need control over a specific character, movement, or rhythm.

Step-by-step order

  1. Choose the "text→video" mode. This is the option where the only source is a description. Leave the image and video upload fields empty — you don't need them at the start.
  2. Write a short, clear prompt. One object, one action, a simple setting. Don't pile up adjectives — clarity matters more than intensity. The prompt is written in English.
  3. Set modest parameters. For the first time: resolution 720p, aspect ratio 16:9, duration 5 seconds, the regular (not Fast) version of the model. That's enough to see the quality without a long wait.
  4. Run the generation. The model will process the text and assemble the clip. Just wait for the result.
  5. Evaluate and adjust. Watch the video. What's off — the object, the movement, the background, the light? Change one or two phrasings in the prompt and run it again. Small, targeted edits give a more predictable result than a full rewrite.

A ready-made prompt for your first try

Copy this prompt as is — it's short, clear to the model, and works great for a beginner. Notice the structure: object and action first, then setting, camera, light, and style.

A red kayak glides across a calm lake at sunrise. Wide shot, slow forward camera move. Soft golden morning light, gentle mist on the water. Cinematic, natural colors, high detail.

Want a different scene? Try: "A barista pours steamed milk into a coffee cup, close-up shot, warm cafe light, shallow depth of field, cinematic." — the same simple formula, a different object.

720p
resolution for the trial
16:9
aspect ratio
5 sec
clip duration
✅ Start cheap
For your first tries use the Seedance 2.0 Fast version and a low resolution (480p or 720p) — that way drafts render faster and cost less. Once you're happy with the composition and movement, regenerate the final version on the regular model at a high resolution (1080p and up). Don't polish something that's still being tested with the expensive version.

What matters about the prompt itself

  • Length limit. The prompt field holds up to 3000 characters. For your first video two or three sentences are plenty — you don't need a long text at the start.
  • The power of the first words. The model gives the greatest weight to the first 20–30 words of the prompt. So put the main object and action right at the beginning, and style details after. A detailed breakdown of the base formula (Subject + Action + Environment + Camera + Lighting + Style) awaits you in a separate lesson on prompt structure.
  • One action per frame. A present-tense verb, one clear movement. "Glides," "pours," "walks" — not three actions at once.
  • Phrase it positively. Describe what you want to see, not what to avoid.

How to evaluate the result

What to checkIf you don't like it — a prompt edit
ObjectRefine the look and color in the first words: "a bright red kayak."
MovementChange the verb or add pace: "glides slowly," "drifts."
CameraSet one shot and one move: "wide shot, slow forward move."
Light and moodDescribe the source and time of day: "soft golden morning light."
StyleAdd a key style word: "cinematic," "documentary."
⚠️ Don't change everything at once
If you rewrite the whole prompt, you won't know which edit actually made the difference. Change one element at a time and compare — that's how you'll quickly learn to feel the model's reaction to your words.

Congratulations — you've completed a full generation cycle: mode, prompt, parameters, run, evaluation, edit. In the next lessons we'll break down the base prompt formula in detail and learn to precisely control camera, light, and style.

Try Seedance 2.0 yourself

Generate your first clip in ZenCreator — Seedance 2.0 is available in the Image-to-Video tool.

Start Creating
PreviousNext lesson