Make a scene from words
This mode is for ideas you can describe before you have assets. Write a person walking into frame, a product turning on a table, or a city street at night, then let Seedance create the first moving version.
Turn a written scene into a video. Describe the subject, movement, camera, and mood, then choose a Seedance model to generate it.
Fill in the form on the left and click "Generate" to create your own video.
Download links expire in 24h — please save your videos promptly.
View My Video History →Type it. Watch it move.
Use text to video when you have an idea but no footage yet. Describe the scene in normal language, choose a Seedance model, and generate a short clip you can watch, compare, and revise.
This mode is for ideas you can describe before you have assets. Write a person walking into frame, a product turning on a table, or a city street at night, then let Seedance create the first moving version.
It helps you test tone before production. Ask for realistic, anime, commercial, documentary, product-demo, or cinematic looks, then keep the direction that feels closest to the final edit.
A good prompt can include camera movement, pacing, mood, and optional sound cues. Mention rain, traffic, footsteps, room tone, or silence when audio helps the viewer understand the moment.
Use it to see whether an idea is worth developing. If the first clip has the right hook, you can refine the prompt, add references, or rebuild the scene as a stronger campaign asset.
Prompt
FORMAT: 16:9 VIDEO CONCEPT: "Hidden Rooftop Break" A highly realistic self-recorded backstage vlog captured entirely by a Korean female idol using a consumer camcorder. The footage should feel genuinely accidental and unscripted, like personal footage never intended for broadcast. Natural hand movement, realistic autofocus behavior, occasional slight framing mistakes, authentic walking motion, believable exposure adaptation, and smooth physical interactions with the environment. No glitches, no duplicated limbs, no warped objects, no facial distortion, no sudden appearance/disappearance of people or items. VISUAL STYLE Early digital camcorder realism. Soft detail, mild sensor noise, subtle compression artifacts, natural skin texture, realistic indoor-to-outdoor exposure transitions, accurate lighting physics, authentic motion blur. No cinematic effects, no artificial lens flares, no beauty filters, no exaggerated color grading. CHARACTER HANA, Korean female idol in her twenties. Long dark hair, natural makeup, clear skin, expressive eyes, slim figure. Wearing a comfortable oversized zip-up hoodie over her stage outfit, in-ear monitors around her neck. LOCATION Upper maintenance level of a large music-show building. Emergency stairwells, rooftop access doors, ventilation units, concrete walls, safety railings, distant city skyline, evening sunlight. Occasional staff members pass naturally in the background. The recording starts mid-walk inside a quiet stairwell. HANA pushes open a rooftop access door and steps outside into the evening air. She immediately smiles at the camera. "I come up here whenever I need two minutes of peace." She walks toward the railing overlooking the city. Wind lightly moves her hair and clothing. The camera naturally turns to reveal the skyline, traffic below, and the glowing broadcast building behind her. "Most people downstairs don't even know this spot exists." She leans casually against the railing while watching the sunset. For several seconds there is no dialogue. Only distant traffic, faint wind, and venue sounds below. She notices a small airplane crossing the sky and instinctively points the camera toward it. "I always get distracted up here." A vibration from her phone in her pocket causes her to laugh. She checks the message briefly. "Yep... they're looking for me." She starts walking back toward the rooftop door. Before entering, she turns around one last time and records the sunset. "Five minutes of freedom. That's all I needed." She opens the door and re-enters the building. The atmosphere immediately changes from quiet rooftop ambience to busy backstage noise. Crew members move through the hallway ahead. Someone calls her name from off-screen. She laughs and begins walking faster. "Okay, okay, I'm coming!" The recording continues naturally while she heads down the stairwell, then ends abruptly as if she simply stopped recording and put the camera away. Important generation requirements: one continuous realistic recording, consistent identity throughout the entire video, stable anatomy, realistic walking mechanics, accurate hand interactions, natural environmental audio, physically correct lighting, no visual artifacts, no AI glitches, no jump cuts, no impossible movements, no scene changes, no duplicated objects or people.
The flow should feel simple: type the idea, choose basic settings, generate the clip, and adjust the part that did not work.
Start with one visible event. Say who is on screen, what happens, and where it happens. A courier checks a package under a shop awning is clearer than a beautiful cinematic city.
Pick a ratio, duration, model, and camera direction. For a first pass, one camera move is enough: locked shot, slow push, gentle pan, or handheld follow.
Watch the first result once for the main action and once for details. If the clip feels rushed, reduce the action. If the scene looks wrong, rewrite only the setting line.
A strong text to video prompt tells the model what should be seen first. Keep it close to a shot direction, not a mood board.
Name the action before the style. The viewer should understand what changes on screen: a chef places a bowl down, a car passes a storefront, or a product rotates under soft light.
Add only details that matter. For this mode, camera, setting, mood, and sound are useful when they protect the idea. Too many competing requests make the clip harder to control.
Revise one thing at a time. Keep the text to video lines that worked, then change the action, setting, or camera note that caused the problem.
Use it when speed and clear direction matter more than exact asset matching. It is a good first step before adding images, references, captions, or final edits.
Use it for quick posts, hooks, reels, and short loops. Start with one action that reads on a phone, then add captions or music after the clip is approved.
This mode can test an ad idea before a shoot. Try a product reveal, a lifestyle scene, a problem-solution opener, or a simple brand mood without booking a crew.
Use text to video to preview a scene, camera move, or character entrance. It helps teams see pacing and framing before they commit to a full storyboard or edit.
Text to video can turn a topic into a visual scene: a process, a product benefit, a classroom example, or a simple metaphor. Add accurate labels and narration in editing.
Short answers before you spend credits on a text to video generation.
Text to video turns a written prompt into a short video. Describe the subject, action, setting, camera, style, and optional sound, then Seedance generates a clip you can review.
Start with one visible action. Add the place, camera move, mood, and sound only when they matter. Good text to video prompts read like short shot directions.
Choose Seedance 2.0 when you want a stronger final pass. Choose Fast or Mini when you want more text to video drafts for the same idea before spending more credits.
Short clips are easier to direct. For text to video, five to eight seconds is a good first test. Use longer durations only when the action is simple and continuous.
Supported Seedance modes can generate audio with the video. Describe a recognizable sound source, such as rain, traffic, footsteps, music, or silence, then review the timing.
Yes, text to video is useful for ad concepts, hooks, mood tests, and rough campaign scenes. Use image or reference inputs later when packaging, labels, or logos must match.
The text to video prompt may contain too many events or camera requests. Remove one detail, put actions in order, and choose one main camera behavior.
Commercial use depends on your plan, provider terms, and rights behind the prompt or references. Use text to video with material you own or can legally use.
Start with one clear action, choose a Seedance model, and generate your next text to video draft from the tool at the top of this page.
Create text to video