Video and audio are generated in a unified forward pass. Wind, footsteps, voice, music — every sound lands on the correct frame because the model never separates them.
Drop a Reference Image for Seedance 2.0
The model accepts up to 9 reference images per generation. Upload a character sheet, a location photo, or a lighting reference — it pulls identity, style, and staging from what you drop in, then directs the scene around it.
Supports PNG, JPG, WebP up to 24MB
Pick Your Aspect Ratio
16:9 for cinematic, 9:16 for shorts, 1:1 for social. Every ratio renders at up to 4K.
Native Audio-Video in One Pass
It is the first mainstream video model that generates audio and video jointly — not as a post-processed layer. Footsteps land on puddles at frame-accurate timing, cloth rustles when the wind blows, a guitar string vibrates in sync with the note.
Prompt
Person walking through puddles in heavy rain, footsteps synchronized with splashing sounds, raindrops hitting umbrella in rhythm with audio, 4K quality, realistic water physics, cinematic atmosphere, perfect audio-visual timing.
Native audio sync
Director-Level Camera Control
Seedance 2.0 takes cinematographer vocabulary literally. Call a dolly-in, a rack focus, a Dutch angle, a whip pan — it executes. Multi-shot storytelling from a single prompt, so one 15-second render can feel like a cut sequence.
Prompt
Professional portrait of a young man in a rainy urban street at night, neon signs reflecting on wet pavement, atmospheric fog, shallow depth of field, cinematic bokeh, moody color palette, 4K ultra-detailed, film noir aesthetic.
Cinematic control
Phoneme-Level Lip-Sync in 8+ Languages
Drop a character portrait and a line of dialogue — the model animates mouth shapes at the phoneme level, not the word level. The result passes on close inspection in English, Mandarin, Japanese, Korean, Spanish, French, German, and more.
Prompt
Close-up shot of a woman speaking directly to camera, clear articulation of words, natural facial expressions during speech, perfect lip-sync with audio, 4K cinematic quality, professional interview lighting, authentic conversational tone.
Phoneme lip-sync
Physics That Hold Up
Fabric wrinkles the way cloth wrinkles. Liquids refract. Particles obey gravity and wind independently. Trained on real-world footage, its world model survives slow-motion scrutiny that kills other video models.
Prompt
Slow-motion shot of a red silk scarf being thrown into the air, floating gracefully with realistic fabric physics, gentle wind affecting movement, 4K quality, cinematic lighting with soft shadows, photorealistic material properties.
Real-world physics
9 Images + 3 Videos + 3 Audios per Generation
Seedance 2.0 takes richer reference payloads than any other public video model. Feed character sheets, location plates, existing footage, reference scores — it fuses them into a single coherent render instead of averaging them into mush.
Prompt
4K close-up of water being poured into a crystal glass, realistic liquid physics with surface tension, light refraction through water and glass, dynamic splashing, photorealistic transparency and reflections, cinematic lighting.
Multi-reference fusion
Topped Artificial Analysis in 2026
It hit Elo 1269 on Artificial Analysis's video-generation leaderboard in April 2026, ahead of Google Veo 3, OpenAI Sora 2, and Runway Gen-4.5. On SeedVideoBench-2.0 it leads text-to-video, image-to-video, and multimodal tasks.
Prompt
Cherry blossom petals falling in slow motion, realistic wind patterns affecting each petal differently, natural gravity and air resistance, 4K cinematic quality, soft bokeh background, spring atmosphere, photorealistic textures.
Benchmark leader
Why Creators Choose Seedance 2.0
Seedance 2.0 dropped the tradeoffs. Cinematic output, native audio, director-level control, and world-class physics — all in one model, in one pass, in one 15-second clip.
Seedance 2.0 Credit Plans
Every render costs credits based on duration and resolution. Pick the plan that matches how much you shoot. Credits roll over on subscription plans; one-time packs never expire.
Starter
$9.9/ month
For solo creators testing the model.
Includes:
- 2,950 credits per month
- ~30 renders/month
Creator
$19.9/ month
For working video creators.
Includes:
- 6,500 credits per month
- ~65 renders/month
Studio
$49.9/ month
For agencies running at volume.
Includes:
- 18,000 credits per month
- ~180 renders/month
Seedance 2.0 FAQ
Quick answers about running Seedance 2.0 through our hosted gateway.
01What is Seedance 2.0?
Seedance 2.0 is ByteDance's flagship multimodal video model, released February 2026. It generates up to 15 seconds with native audio, director-level camera control, and phoneme-level lip-sync in 8+ languages, all in a single forward pass. The model topped Artificial Analysis's leaderboard above Veo 3, Sora 2, and Gen-4.5.
02Is Seedance 2.0 free to try?
You get starter credits on sign-up — enough to render your first clip without paying. After that, every render costs credits based on duration and resolution. Our credit plans start at $9.90/month.
03Can I use Seedance 2.0 from the US?
Yes. ByteDance excluded the United States from the direct rollout through Dreamina. Our hosted gateway relays requests through supported regions, so US creators can use the full Seedance 2.0 API without a VPN or waitlist.
04How long are Seedance 2.0 videos?
Each render is up to 15 seconds. Within that window the model can produce multiple shots with natural cuts and transitions, so the output feels like an edited sequence rather than a continuous take.
05What inputs does Seedance 2.0 accept?
In a single pass, Seedance 2.0 accepts a text prompt plus up to 9 reference images, 3 video clips, and 3 audio clips. Character identity, location, camera style, and even ambient sound can all be seeded from references.
06Does Seedance 2.0 really generate audio?
Yes, and it is one of the defining features. Video and audio are generated jointly in one forward pass — not post-processed. Footsteps, dialogue, music, and ambient sound all land on the right frame because the model never separates them.
07Is Seedance 2.0 safe for commercial use?
Renders you generate through our gateway are yours to use commercially under our terms of service. Seedance 2.0 has built-in content moderation; prompts that violate policy are rejected before compute.