Seedance AI Video Generator: Pro and Lite Text to Video
Flagship text-to-video model for rich multi-shot storytelling.
Keep prompts under 10,000 characters. Add scene cuts using [Cut to] markers.
Camera fixed
Lock the camera position to reduce motion drift.
Safety checker
Disable only if you need unrestricted outputs via API.
Use -1 for random results or set a fixed seed for reproducibility.
About Seedance
Seedance is ByteDance's video family, and this page puts four variants in one workspace: V1 Pro and V1 Lite text to video, plus Pro and Lite image to video. Clips run from three to twelve seconds at up to 1080p, prompts can carry scene cuts, and one flat credit charge applies per run before you start.
The reason to open Seedance is control over length. Several models on this site hand you a fixed pair of durations; Seedance accepts any whole number from three to twelve seconds, in one-second steps, at 480p, 720p or 1080p. That matters when a shot has to fit an existing edit: a six-second insert and an eleven-second establishing shot are both possible without generating long and trimming in an editor.
The four variants are not cosmetic. V1 Pro text to video is the flagship and takes the widest ratio range, from 21:9 through 1:1 to 9:16, so one prompt can serve a cinema crop and a vertical feed. V1 Lite text to video trades fine detail for turnaround and adds 9:21. The image modes drop aspect controls because the source sets the frame, and Lite image to video also accepts an optional ending image, which is how you land a shot on a specific final pose.
Prompts here are longer than on the sibling model pages, up to 10,000 characters, and [Cut to] markers split one prompt into several shots. A camera-fixed switch pins the frame when drift would show, and a seed field makes a good result repeatable. Compare that with Sora 2, which caps clips at ten or fifteen seconds and keeps multi-shot planning in a separate storyboard mode, or with Runway, whose Gen-4 tops out at eight seconds.
A new Seedance form is not blank: it opens at 16:9, 720p and five seconds, with the camera fixed switch off and the seed empty. Change the length first, since that is the reason to be on this page, then set the resolution, and only then write the prompt. Leaving the camera fixed switch off lets a described camera move actually happen; turn it on for locked-off shots, where it holds the position and reduces drift, and accept that the move you described will not appear. Settling the frame before the prose saves a rewrite.
An empty seed field is treated as randomised, so two runs of an identical prompt return different takes. Set a number once a take works, and the same prompt, length and resolution will come back close to it. That makes the seed the right control for comparing prompt edits: fix it, change one phrase, run again, and the difference you see belongs to the wording rather than to chance. Run at 720p to judge motion, and switch to 1080p only when the framing will be inspected closely.
How Seedance works
- 1
Pick a variant
Pro text to video for narrative prompts, Lite text to video for quick passes and 9:21, Pro image to video for motion from a finished still, Lite image to video for drafts with an optional end frame.
- 2
Set the frame
Choose three to twelve seconds, then 480p, 720p or 1080p. Text modes take a ratio between 21:9 and 9:21, while image modes inherit the source. Upload or paste a PNG, JPG or WEBP file no larger than 10MB.
- 3
Write the cut list, then generate
Describe motion, lighting and camera moves in under 10,000 characters. Use [Cut to] where the scene changes. Fix the seed if you want to reproduce a take, then run; the credit cost is shown first.
What Seedance is best for
Inserts that fit an edit
A six- or seven-second shot built to a music edit, instead of a ten-second clip you have to cut down before it fits.
Tall 9:21 placements
Lite text to video reaches a taller ratio than most feeds ask for, while Pro keeps 9:16 and 21:9 in the same workspace.
Multi-shot narrative drafts
One prompt with two or three [Cut to] markers sketches a whole sequence before you commit to a larger model.
Motion from product stills
Pro image to video animates a clean packshot, and Lite image to video with an end frame is the fast way to test a transition.
Seedance at a glance
- Model
- ByteDance Seedance V1 Pro and V1 Lite
- Modes
- Pro and Lite text to video, Pro and Lite image to video
- Clip length
- 3 to 12 seconds, in one-second steps
- Resolution and ratio
- 480p, 720p or 1080p; 21:9 to 9:16 on Pro text, 16:9 to 9:21 on Lite text, source frame on image modes
- Prompt and controls
- Up to 10,000 characters with [Cut to] breaks, camera-fixed switch, seed, safety checker, optional end frame on Lite image
- Output
- MP4, downloaded from the preview panel
Seedance FAQ
- Which Seedance variant should I choose?
- Pro text to video for a narrative prompt with several shots, Lite text to video when you want a fast read and the 9:21 ratio, Pro image to video for detailed motion from a finished still, and Lite image to video for draft passes, especially when you need an ending frame.
- How long can a Seedance clip be?
- Any whole number of seconds from three to twelve. There is no fixed pair of durations, so you can match an edit instead of trimming to fit. Resolution is independent of length: 480p, 720p or 1080p.
- Can one prompt contain several shots?
- Yes. The prompt field takes up to 10,000 characters and [Cut to] markers separate scenes. Pro text to video is the variant positioned for multi-shot storytelling, and the same marker convention appears in the other three prompts.
- How do the image-to-video modes take an image?
- Paste a hosted URL or upload a PNG, JPG or WEBP file no larger than 10MB. The image supplies the aspect ratio. Lite image to video also accepts an optional ending image, so the shot can move toward a final frame instead of stopping wherever the model lands.
- What does a run cost, and what happens if it fails?
- One flat credit charge per run, displayed in the generator before the request goes out. If the provider returns no video, the charge is refunded. Plans bundle monthly credits, and Starter, from $4.90 a month, carries 300 of them.
- What does the camera fixed switch do?
- It locks the camera position, which is how you reduce drift on a shot that should not move at all. It starts switched off, so a camera move written into the prompt is allowed to happen; switching it on tells the model to hold the frame instead. Use it for locked-off product shots and talking heads, leave it off for pans, pushes and handheld motion, and change only this one setting between runs so you can tell what it did.