Skip to content

How do I use Seedance?

2026-08-01 · help-centre article · Higgsfield · original

Topic: Video generation & shot direction (also: Prompting & AI assistants, Assets, references & consistency, Failure modes & fixes, Higgsfield platform & ecosystem, Cost, credits & plans, Audio, voice & music) · mode: Reference · tags: flattened table

This help-centre guide explains Seedance, ByteDance's video family on Higgsfield, which makes video with native sound and lip sync from text, images, clips and audio at once. Seedance 2.5 runs up to 30 seconds in any ratio from 9:16 to 21:9 on Pro and Plus plans and above, while Seedance 2.0 makes 4 to 15 seconds. The working method is to start with image-to-video at 720p, write the prompt in four parts (action, setting and light, camera move, mood), then rerun the locked prompt at 1080p or 4K. To keep a character across clips, save it as an Element and call it with @.

What is said

  • Short answers on resolution, length, sound and Elements

What is said

★ — how much it matters to the source.

  • ★★★ Higgsfield describes Seedance as ByteDance's multimodal video family for realistic motion: it takes text, images, video clips and audio at once and makes video with native sound in one pass, with multilingual lip sync; Seedance 2.0 to 2.5 are in the Video model picker. · Seedance, ByteDance, native audio, lip sync, Seedance 2.0, Seedance 2.5, image-to-video
  • ★★ On Higgsfield, Seedance 2.5 is the latest Seedance version, with the longest clips (4 to 30 seconds), synchronised audio and up to 1080p on Pro and Plus plans and above. · Seedance 2.5, Pro plan, Plus plan, Seedance
  • ★★ Seedance 2.5 Edit edits existing clips, audio included, at 480p to 720p, with the output length matching the source clip. · Seedance, Seedance 2.5
  • ★ Older Seedance versions (Seedance 1.5 Pro, Seedance Pro and Seedance Pro Fast) stay available on Higgsfield under All models in the model picker. · Seedance
  • ★★★ Seedance 2.5 generates clips up to 30 seconds long in any aspect ratio from 9:16 to 21:9, with synchronised audio made in the same pass. · Seedance 2.5, native audio
  • ★★ Higgsfield claims Seedance 2.5 follows prompts noticeably better than Seedance 2.0, so structured prompts translate into the shot more reliably. · Seedance 2.5, Seedance 2.0
  • ★★ Seedance 2.5 Edit is a separate model that changes an already generated clip, including region-level edits marked directly on the frame, and its audio. · Seedance, Seedance 2.5
  • ★★★ Seedance 2.5 is available on Higgsfield's Pro and Plus plans and above, not on Starter and Basic. · Seedance 2.5, Pro plan, Plus plan, Starter plan, Basic plan
  • ★★★ For a first Seedance 2.0 clip (Video → Seedance 2.0), Higgsfield advises image-to-video at 720p with a sharp, front-facing, well-lit reference image and a short prompt of action and camera move, since it is more predictable than text-only and costs fewer credits. · Seedance 2.0, image-to-video, credit
  • ★★★ Higgsfield advises a structured Seedance 2.0 prompt covering four things in order: what the subject does, setting and lighting, camera move, mood or atmosphere; vague prompts give generic motion. · Seedance 2.0
  • ★★ Higgsfield advises reviewing the whole Seedance 2.0 clip before iterating, since most quality problems appear at the 5 to 8 second mark: check that the motion holds and the camera move happened. · Seedance 2.0
  • ★★ When a Seedance 2.0 clip is off, Higgsfield advises changing one thing per iteration: the prompt or the reference, not both. · Seedance 2.0 · first in: 2026-06-24, Generating with Seedance 2.0: Full Tutorial
  • ★★★ Once a Seedance 2.0 prompt gives the right composition at 720p, Higgsfield advises running the identical prompt at 1080p or 4K. · Seedance 2.0
  • ★★ The Seedance 2.0 generation bar on Higgsfield sets clip duration, aspect ratio (with Auto inferring it from the input), resolution up to 4K and bitrate; the credit cost on the Generate button updates as they change. · Seedance 2.0, credit
  • ★★ Higgsfield lists camera terms Seedance reads directly: dolly in, truck left, arc shot, push in, pull back wide, handheld follow, crane up, orbital move. · Seedance · first in: 2026-06-24, Generating with Seedance 2.0: Full Tutorial
  • ★★ Higgsfield's example Seedance prompt: a character walking through a train station at rush hour with a close shoulder-height tracking camera, warm fluorescent light, a tense mood and an "SFX:" line for ambient noise. · Seedance
  • ★★ Seedance's references on Higgsfield: @character holds face geometry, skin tone and style within a clip, @style applies lighting, palette and mood, @motion copies camera behaviour and motion from a clip, @audio syncs visuals and lip sync to audio. · Seedance, lip sync
  • ★★★ One Seedance 2.0 generation combines up to 9 images, 3 video clips and 3 audio files; one Seedance 2.5 generation up to 30 images, 10 video clips and 10 audio files. · Seedance 2.0, Seedance 2.5, image-to-video
  • ★★★ For the same character across separate Seedance generations, Higgsfield advises saving it as an Element and calling it with @; for a real person, train a Soul ID first and build the Element from its portraits. · Seedance, Elements, Soul ID
  • ★★ When Seedance ignores the camera instruction, Higgsfield advises replacing vague terms with specific ones: dolly in, pull back wide, arc right, crane up. · Seedance
  • ★★ When a Seedance character changes halfway through a clip, Higgsfield advises adding visual anchors (hair colour, clothing details, distinguishing features) and referencing the character as an Element across generations. · Seedance, Elements
  • ★★★ For jerky motion at a cut between Seedance clips, Higgsfield advises generating the last frame of one clip and the first frame of the next, then making the transition between them with first-and-last-frame. · Seedance, keyframe · first in: 2026-06-29, How to Keep Your AI Character Consistent Across Every Generation: Soul ID Explained
  • ★★★ Since resolution and clip duration drive Seedance credit use, Higgsfield advises prototyping at 720p with short clips, locking the prompt, then running the final at full resolution and length. · Seedance, credit
  • ★★ To use Seedance in Higgsfield Canvas, connect a reference to the Seedance node and describe the reference's role at the start of the prompt. · Seedance, Canvas
  • ★★★ Seedance generations through Canvas, MCP, the CLI and Supercomputer always deduct credits, whatever Unlimited access the user has on the web. · Seedance, Canvas, Higgsfield MCP, Supercomputer, Higgsfield Unlimited, credit
  • ★★★ A single Seedance generation cannot run longer than 30 seconds, the Seedance 2.5 maximum; Seedance 2.0 generates 4 to 15 seconds. · Seedance 2.5, Seedance 2.0, Seedance
  • ★★★ Turning Seedance's sound on or off does not change a generation's cost on Higgsfield: credits depend on model, resolution and clip duration, shown on the Generate button. · Seedance, native audio, credit

Threads in this video

Where the stories that moved here went next

Mentioned

Central: Seedance · Seedance 2.0 · Seedance 2.5

Also: Organizations: ByteDance

Concepts: credit · image-to-video · keyframe · lip sync · native audio

Products: Basic plan · Canvas · Elements · Higgsfield MCP · Higgsfield Unlimited · Plus plan · Pro plan · Region Edit · Seedance 1.5 · Seedance 2.5 Edit · Soul ID · Starter plan · Supercomputer

← Previous: What external integrations does Higgsfield support? · Next → What should I do if my payment failed?