Seedance 2.0 AI Video Generator
Seedance 2.0 is ByteDance's video model, and it runs right here. Describe a shot, animate a still, restyle a clip you already have, or hand it images, video and audio as references.
- Free credits on signup
- No credit card to start
- No watermark
- Nothing to install
Seedance 2.0 at a glance
- Clip length
- 4-15s
- Max resolution
- 1080p
- Frame shapes
- 7
- References per shot
- Up to 15
Seedance 2.0 video generator
What Seedance 2.0 Can Do
The controls that matter, and what each one changes about the clip you get back.
Four Seedance 2.0 Modes in One Place
Start from a sentence, a still image, an existing clip, or a pile of references. All four run on the same model, so you are not relearning a new tool every time the material changes.
Clips From 4 to 15 Seconds
Five seconds is the default and covers a single beat. Push it to fifteen when an action needs room to finish — a door opening, a car passing, a line of dialogue landing.
Up to 15 References Per Shot
Nine images, three video clips and three audio files, in any combination. Wardrobe can come from one photo, the location from another, the pacing from a clip you already like.
First and Last Frame Control
Pin the opening frame, the closing frame, or both, and Seedance 2.0 fills the motion in between. It is the difference between describing an ending and actually specifying one.
Audio Made With the Picture
Switch audio on and the sound is produced in the same pass as the image, so footsteps land on the footfall instead of drifting a few frames off.
Seven Shapes, Three Resolutions
1:1, 4:3, 3:4, 16:9, 9:16, 21:9 or adaptive, at 480p, 720p or 1080p. Choose the frame the platform wants before you generate rather than cropping it back afterwards.
How to Use Seedance 2.0
Three steps to your first clip:
- 01
Pick the Mode That Matches Your Material
Nothing but an idea? Stay on text to video. Got a still you like? Image to video. Got footage, or a character you need to keep consistent? Video to video and reference to video are the two tabs for that.
- 02
Write the Shot, Then Attach References
One action, one camera move, one lighting note. Then upload whatever should stay fixed — a face, a product, a location — instead of trying to describe it in words.
- 03
Generate, Then Change One Thing
Set duration, shape and resolution, and generate. If the result is close but not right, change a single variable and run it again — that is how you learn what Seedance 2.0 responds to.
What People Make With Seedance 2.0
Four jobs it handles without a camera, a crew, or a booked studio day.
Short Ads Shot End to End
Fifteen seconds is a whole pre-roll. Write the beat, pin the product shot as a reference, and get a spot you can cut straight into a campaign.
- Product stays on model — Upload the packshot as a reference so the label does not drift between takes.
- Cut-ready shapes — 21:9 for pre-roll, 9:16 for the same spot on phones.
Vertical Social Clips
Shorts, Reels and TikTok all want 9:16 and they all want it now. Generate in the shape the platform expects instead of cropping a wide frame down to it.
- Several takes cheap — Run 480p while you are still deciding, then re-run the winner at 1080p.
- Sound included — Turn on audio and the clip is postable without a second pass.
Characters That Survive a Cut
The reason most AI scenes fall apart is that the person changes between shots. Feed Seedance 2.0 the same character images every time and the face, hair and wardrobe hold.
- Nine reference images — Enough for a face, a wardrobe and a couple of angles.
- Reference clips too — Three short videos can carry motion and pacing, not just looks.
Product Shots That Move
A still packshot becomes a slow orbit or a push-in. Pin the first frame to your existing photo so the motion starts from the asset you already signed off.
- Starts from your photo — First-frame control means the product is yours, not a lookalike.
- One move per clip — Orbit or push-in, not both — mixed moves are where shots go soft.
How to Write Seedance 2.0 Prompts
Video prompts fail differently from image prompts. The usual cause is not vagueness — it is asking for three events in a five second clip.
A shot description covers five things
- 1Subject — who or what is in frame, and the one thing they are doing
- 2Camera — a single move: tracking, push-in, orbit, static
- 3Lighting — neon, overcast, golden hour, single practical lamp
- 4Setting — where it happens and what is behind the subject
- 5Pace — slow and held, or quick and handheld
a woman walking in the city
A woman in a red raincoat walks past a noodle shop, camera tracking alongside at shoulder height, neon signs reflecting in wet pavement, slow steady pace.
product video, cool and modern
A matte black bottle on a concrete slab, camera orbits ninety degrees to the right, hard side light from a single softbox, shallow depth of field.
he opens the door, walks out, gets in the car and drives away
A man pushes a warehouse door open and steps into daylight, static wide shot, dust in the air, he stops just past the threshold.
Seedance 2.0 FAQ
It is ByteDance's video generation model. It takes a written shot description and optional reference material — images, clips, audio — and renders a short video with camera motion, and sound if you ask for it. On this page it runs in four modes: text to video, image to video, video to video and reference to video.
New accounts are granted free credits, so you can start generating before paying anything. It is not unlimited — every generation spends credits, and what a single clip costs depends on its length and resolution (see the next answer). No card is needed to sign up.
Cost is per second of output, scaled by resolution — a second at 720p costs more than at 480p, and 1080p more again. The exact figure sits on the generate button and updates as you change resolution or duration, so you always see it before you spend anything.
Between 4 and 15 seconds, set with the duration control. It defaults to 5, which is about right for one action. Fifteen is the ceiling for a single generation — longer pieces are built by generating several shots and cutting them together.
480p, 720p and 1080p, with 720p as the default. Seven frame shapes are available: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, and adaptive, which follows the shape of your input image. Output has no watermark at any resolution.
Only if you switch audio on — it is off by default. When it is on, the sound is generated in the same pass as the picture rather than added afterwards, so effects line up with what is happening on screen. You can also upload up to three audio files as references.
Fast bills at a lower base rate and turns clips around quicker, which makes it the sensible choice while you are still testing prompts. Move to the standard model once you know the shot you want and you are generating the version you will actually ship.
Yes. Anything you can generate on this page can be triggered from our REST API with an API key from your dashboard, and it bills from the same credit balance. The model id follows the same scene-and-model form the studio uses.
Nearly always because the prompt asked for more than one event. A five second clip holds one action; if you describe someone standing up, crossing a room and opening a door, the model picks. Cut it back to a single beat, name one camera move, and attach a reference image for anything that has to look exactly right.
Make Your First Seedance 2.0 Video
No camera, no crew, no card. Describe a shot, pick your mode, and your signup credits get you generating.