Seedance 1.5 Pro
Seedance 1.5 Pro from ByteDance generates video and its soundtrack together, including lip-synced dialogue in six languages, from a prompt or from first and last frames, in clips of 4 to 12 seconds.
Overview
Seedance 1.5 Pro is ByteDance Seed's model for generating picture and sound in one pass. Its technical report describes a dual-branch Diffusion Transformer joined by a cross-modal module, trained with supervised fine-tuning and then RLHF, with tight audio-visual sync as the goal.
ByteDance lists ambient sound, action effects, synthesized audio, instruments, background music and vocals among the sounds it produces. Dialogue can be a monologue or a multi-character conversation in Mandarin, English, Japanese, Korean, Spanish or Indonesian, plus Cantonese, Sichuan and Shaanxi dialects and a Taiwanese accent, with lip movement matched to each speaker.
Choosing a variant: use this one when a clip needs talking characters or a finished soundtrack. Seedance 1.0 Pro is the silent predecessor, offered here at 480p or 1080p in clips of 5 or 10 seconds. Seedance 2.0 R2V and its lighter sibling Seedance 2.0 Fast R2V require a reference video and copy its motion or camera work into a new clip.
What you can build
- Short drama scenes with two or more speakers, cutting between shot and reverse shot.
- Product spots with a voiceover in a described voice, tone and pace.
- Localized social clips, made by changing the dialogue language in the prompt.
- Music-led pieces, such as a character clapping on the drum beat of the background track.
- Talking portraits: a still photo as the first frame, with motion and speech added.
Inputs
- A prompt, written in Chinese or English. ByteDance's formula is subject, movement, environment, camera movement, aesthetics and sound, and only the first two are required. Put spoken lines in double quotes and name the speaker.
- An optional first frame in HEIC, JPG, PNG, WebP or GIF. A last frame can be added too, but it is ignored unless a first frame is set.
- Resolution: 480p (default), 720p or 1080p.
- Ratio: adaptive (default), or one of six fixed shapes from 21:9 widescreen to 9:16 vertical.
- Duration: any whole number of seconds from 4 to 12, default 5.
- Generate audio, on by default.
- A watermark toggle and a seed (-1 for random).
Outputs
An MP4 at 24 fps. With audio on, the soundtrack is mono and generated to fit the picture, and ByteDance says background music matching the prompt is added by default. With audio off, the clip is silent.
Recommended settings
- Describe each voice by emotion, tone and speaking pace. ByteDance's examples hold one timbre while the delivery moves from calm to restrained.
- Give every character fixed traits such as gender, age and clothing, so lines land on the right face.
- Write cuts in sequence, for instance "the shot cuts to a close-up of the boy".
- Prepare first and last frames at the same shape. When they differ, the first frame sets the format and the last one is cropped to fit.
Limitations
- A static shot has to be requested in the prompt, and ByteDance does not promise one.
- No reference video or audio input: motion and voice come from the prompt and frames alone.
Safety & compliance
A photo of a real person used as a frame can be made to speak, so use one only with that person's consent.
Example prompts
Great starting points for seedance-pro-v1.5.
API quick start
Run seedance-pro-v1.5 with a single API call.
{
"prompt": "A bright, sunny morning in a modern metro…",
"resolution": "480p",
"ratio": "adaptive",
"duration": 5
}curl -X POST "https://api.wiro.ai/v1/Run/bytedance/seedance-pro-v1-5" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "A bright, sunny morning in a modern metro…",
"resolution": "480p",
"ratio": "adaptive",
"duration": 5
}
JSON