Guide
Seedance 2.0 on OpenClaw: ByteDance Video for Your Agent
Seedance 2.0 is ByteDance's current video family — native audio, multi-shot editing, real-world physics, and director-level camera control, in text, image and reference modes. It sits in your agent's video model dropdown behind a fal key. Here is how the three speed tiers differ and what they really cost.
Seedance 2.0 vs the Seedance You May Already Know
If you have used Seedance on OpenClaw before, it was probably Seedance 1.5 Pro through a BytePlus key. Seedance 2.0 is the next generation and a different route: it is served through fal.ai, so a fal key is what unlocks it. The older 1.5 Pro and 1.0 Pro endpoints are still on fal too if you want them.
What 2.0 adds over the 1.x line is the multi-modal front door. Alongside plain text-to-video and image-to-video, it takes a reference-to-video mode that accepts up to 9 images, 3 video clips, and 3 audio clips in one request — which is what lets an agent hold a character, a location, and a voice steady across a series of clips instead of re-rolling the look every time.
Three Modes
- Text to video — cinematic output from a prompt, with native audio, multi-shot editing and camera direction.
- Image to video — animates a still, with start-and-end-frame control and motion prompting.
- Reference to video — up to 9 images, 3 videos and 3 audio clips as combined references, with native audio out.
Three Speed Tiers, Very Different Prices
| Tier | 480p | 720p | 1080p |
|---|---|---|---|
| Seedance 2.0 | — | $0.3034 / sec | $0.682 / sec |
| Seedance 2.0 Fast | — | $0.2419 / sec | — |
| Seedance 2.0 Mini | ~$0.0721 / sec | ~$0.1547 / sec | — |
Underneath, fal bills Seedance on video tokens rather than wall-clock seconds — tokens are derived from height x width x FPS x duration, so the per-second figures above are the practical translation at each resolution. The headline consequence is that resolution, not duration alone, drives your bill: 1080p on the standard tier is more than double 720p.
Setting It Up
- Create a fal key at fal.ai/dashboard/keys and save it on your API Keys page.
- Open the Video dropdown on your instance card and select the fal.ai provider group.
- Pick a Seedance row — or type “Seedance” in the dropdown search box, since the fal list is refreshed daily from trending endpoints and the visible set shifts.
- Ask your agent for a clip in chat and the MP4 comes back in the conversation.
Model IDs for Self-Hosted OpenClaw
The video model id is the fal endpoint path with a fal/ prefix. Note that the Seedance 2.0 endpoints sit under a bare bytedance/ namespace, while the older Pro endpoints carry the fal-ai/ prefix:
fal/bytedance/seedance-2.0/text-to-video
fal/bytedance/seedance-2.0/image-to-video
fal/bytedance/seedance-2.0/reference-to-video
fal/bytedance/seedance-2.0/fast/text-to-video
fal/bytedance/seedance-2.0/fast/image-to-video
fal/bytedance/seedance-2.0/fast/reference-to-video
fal/bytedance/seedance-2.0/mini/image-to-video
fal/bytedance/seedance-2.0/mini/reference-to-video
# previous generation, still available
fal/fal-ai/bytedance/seedance/v1.5/pro/image-to-video
fal/fal-ai/bytedance/seedance/v1/pro/image-to-videoSeedance 2.0 or Kling or H3?
All three are in the same dropdown, so the choice is about fit rather than access:
- Seedance 2.0 for physics and camera direction, and for the cheapest usable tier in the picker (Mini at 480p).
- Kling for the widest spread of price points and explicit start/end frame control via O3.
- MiniMax H3 for 2K output and reference-driven continuity with audio references.
What's Next?
- fal.ai on OpenClaw — the provider setup behind Seedance 2.0
- BytePlus + Seedance 1.5 — the direct ByteDance route for the previous generation
- Make videos with your bot — the end-to-end brief-to-MP4 flow
- All supported models — chat, image and video across every provider