// ENGINE 02 — BYTEDANCE
Seedance 2.0
Next-gen cinematic video with audio. Cinematic all-rounder — native audio, and the only engine with first/last/reference image control.
- CREDITS
- 60 cr
- QUALITY
- best
- SPEED
- fast
- DURATIONS
- 5 / 10 / 15s
- ASPECTS
- 9:16 · 16:9 · 1:1 · auto
- IMAGE INPUT
- first / last / ref
GENERATE
— renders with seedance 2.0 — 60 crPROMPT SAMPLES
— written for this engine — use one as a starting pointPROMPT_01
A lighthouse keeper climbs a spiral staircase at dusk, lantern light sweeping the stone walls; at the top he opens the hatch and wind rushes in — waves and gulls audible below.
PROMPT_02
Macro shot of an espresso being pulled: crema swirling, steam curling in backlight, slow camera drift right, rich café ambience.
PROMPT_03
A paper boat drifts down a rain gutter through a neon-lit city at night, camera tracking low at water level, rain hissing on pavement.
PROMPT_04
Continuous shot: a dancer in a red dress spins in an empty warehouse, dust catching shafts of sunlight, her footsteps echoing in the space.
PROMPT_05
Fifteen seconds, three beats: a seed sprouts in macro, the sprout becomes a sapling in time-lapse, the camera pulls back to reveal a full oak on a hillside at sunset.
PROMPT_06
A vinyl record drops onto a turntable, the needle settles with a soft crackle, and the camera orbits the spinning label as warm jazz fades in.
ABOUT
Seedance 2.0 is ByteDance's top-tier video engine and the strongest all-rounder in the studio: cinematic image quality, coherent multi-beat motion, and native audio in every render. It holds subjects together across the full 15-second window without the identity drift that plagues long generations.
It is also the only model with all three image modes. Start a video exactly on an uploaded frame, pin the ending with a first + last pair for morphs and transitions, or hand it a reference image and let it keep that character or product consistent through the whole shot. No other engine in the studio offers last-frame or reference control at all.
Despite sitting at the top of the quality tier, Seedance renders fast, and it frames in all four aspect modes — 9:16, 16:9, 1:1, or auto. At 60 credits it is a premium spend that works the first time. When a shot needs more than 15 seconds, step up to Seedance 2.5; to iterate cheaply first, draft on P-Video.
STRENGTHS
— what this engine does better than the others- 15-second shots with subjects that stay coherent to the last frame
- Native audio generated with the picture: ambience, effects and voices arrive in sync
- First + last frame mode interpolates between two images for morphs and transitions
- Reference mode keeps a face, mascot or product consistent across renders
- Top-tier image quality at fast render speed, in every aspect ratio
PROMPTING TIPS
— how to talk to this model- TIP_01Structure long prompts as beats: what happens in the first, middle and final seconds. Seedance follows sequence better than any other engine here.
- TIP_02Name the sounds you want — "waves and gulls audible below" — and the audio pass will include them instead of guessing.
- TIP_03For first + last mode, keep both images in the same style and lighting; the morph quality depends on how plausible the in-between is.
- TIP_04In reference mode, describe the action, not the subject — the image already defines who; the prompt should define what they do.
USE CASES
01
Short-film beats
Multi-second story moments — an establishing shot that develops, a reveal, a slow push-in — that need the full 15 seconds to land.
02
Frame-to-frame morphs
First + last mode interpolates between two uploaded images: product transformations, before/after reveals, seamless scene transitions.
03
Consistent characters
Reference mode keeps a face, mascot, or product identical across renders — the closest thing here to a reusable cast member.
04
Sound-on social video
Audio is generated with the picture, so ambience and effects arrive synced — no separate sound pass before posting.