How to Use Seedance 2.0 in CapCut & Dreamina

Seedance 2.0 isn't a standalone app — inside ByteDance's ecosystem it's the model that powers Dreamina (the international brand) and Jimeng / 即梦 (the China version), and it's wired into CapCut's AI video tools. Here's how to actually generate with it, including the video-reference feature people keep asking about.

Where to find Seedance 2.0

• Dreamina (dreamina.capcut.com) — the most direct route internationally. Sign in with email or Google.
• CapCut — Seedance 2.0 shows up inside CapCut's AI Video and Video Studio features (the AI generation there is driven by Dreamina).
• Jimeng / 即梦 — the China-facing version, same workflow.

Step-by-step (Dreamina)

1. Open Dreamina and sign in.
2. Click AI Video.
3. Choose the Seedance 2.0 model.
4. Pick a mode:
   • Single-frame — supply a first (or last) frame to guide the shot.
   • Multiframes — the multimodal mode where you upload images, video, and audio as references.
5. Write your prompt, and use @AssetName to tell the model what each upload is for (e.g. "@video1 for the camera movement", "@image1 as the first frame").
6. Hit Generate (typically ~60–90 seconds per clip), then Download — output is up to 1080p.

The video-reference feature explained

This is Seedance 2.0's headline capability. In one generation you can feed the model a mix of references — up to 9 images + 3 videos + 3 audio clips (each video/audio ≤15s):

• Video references let the model copy the camera movement, action/choreography, visual effects, or editing rhythm of a clip you upload — or continue/extend footage you already have.
• Image references set characters, first/last frames, or style.
• Audio references drive sound and, combined with a character, lip-synced dialogue.

The @AssetName syntax is what makes this precise: instead of hoping the model guesses, you assign each asset a job in the prompt.

Two practical notes: outside China, use Dreamina international or CapCut — Jimeng requires a Chinese phone number. And during rollout, uploads containing real, recognizable human faces are blocked by content policy, so reference footage of real people may be rejected. For all the access routes, see how to access Seedance 2.0.

If the app route is a hassle

Regional gating, paid-only tiers, and staged rollouts mean not everyone can open Dreamina and start immediately. If you just want to generate a Seedance clip without wrangling accounts, start from a ready-made prompt and run it on ponpon, which runs the Seedance model line for you. To animate talking characters, see Seedance lip sync.