Seedance 1.0 Pro
Seedance 1.0 Pro from ByteDance makes silent 5 or 10 second clips at 480p or 1080p from a prompt, a first frame or a first and last frame pair, and can cut between several shots inside one prompt.
Overview
Seedance 1.0 Pro is the large-parameter model in ByteDance's Seedance 1.0 series, trained jointly on text-to-video and image-to-video. Multi-shot output is built in: one prompt can describe several shots, and ByteDance says the main subject, visual style and atmosphere carry across the cuts. ByteDance also reports stable motion from subtle facial expressions up to large-scale movement.
Choosing a variant: pick this one for silent footage that you will score or voice yourself. Seedance 1.5 Pro adds generated dialogue, effects and music to the same prompt-and-frames workflow, along with a 720p setting and any length from 4 to 12 seconds. Seedance 2.0 R2V and Seedance 2.0 Fast R2V start from reference videos, for when an existing clip's movement should carry into a new scene.
What you can build
- Animated product shots from a single packshot used as the first frame.
- Transitions between two keyframes, such as a before and after state.
- Short narrative scenes with two or three camera switches in one prompt.
- Stylized clips in looks ByteDance shows for text-to-video, including clay, felt, voxel, pixel art, line drawing, manga and American comic.
- Longer sequences built clip by clip: take the last frame of one result and upload it as the first frame of the next.
Inputs
- A prompt in English or Chinese. ByteDance suggests subject plus action for simple shots, actions in the order they happen for longer ones, and "Camera switch." between shots.
- An optional first frame image, and an optional last frame that is used only together with a first frame (JPG, PNG, WebP, GIF or HEIC).
- Resolution: 480p (default) or 1080p.
- Aspect ratio: adaptive by default, or 16:9, 4:3, 1:1, 3:4, 9:16 or 21:9.
- Duration: 5 seconds (default) or 10 seconds.
- To keep the camera still, say so in the prompt, though ByteDance does not guarantee a static result.
- A watermark switch (off by default) and a seed, with -1 for random.
Outputs
One MP4 at 24 frames per second, without sound. A 16:9 clip is 1920×1088 in 1080p and 864×480 in 480p. The optional watermark is an "AI Generated" label in the lower-right corner.
Recommended settings
- For image-to-video, upload the sharpest source you have; ByteDance says input quality has a large effect on the result.
- With a first frame, match the ratio to the image or leave it on adaptive, because a mismatch crops the image around its center. For prompt-only runs, choose a fixed ratio, since ByteDance supports adaptive on this model only when an image is given.
- Name camera moves directly: push, pull, pan, truck, orbit, follow, rise or zoom. They can be chained into one long take.
- Prompt-only results change a lot from one seed to another, and the seed stays at 1 unless you change it. Try a few fixed numbers and note the one behind a clip you like, since a -1 (random) run cannot be repeated. The same seed gives a similar clip, though ByteDance makes no promise of an exact match.
Limitations
- No audio generation.
- Uploaded images that fall outside ByteDance's limits (sides of 300 to 6000 pixels, 30 MB, aspect ratio 0.4 to 2.5) are resized, compressed or center-cropped to fit, so a very wide or tall image loses its edges.
Safety & compliance
Use photos of real people as first or last frames only with their consent.
Example prompts
Great starting points for seedance-pro-v1.
API quick start
Run seedance-pro-v1 with a single API call.
{
"prompt": "A Red Bull Formula 1 car roars into a sha…",
"resolution": "480p",
"ratio": "adaptive",
"duration": 5
}curl -X POST "https://api.wiro.ai/v1/Run/bytedance/seedance-pro-v1" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "A Red Bull Formula 1 car roars into a sha…",
"resolution": "480p",
"ratio": "adaptive",
"duration": 5
}
JSON