ByteDance’s cloud arm has put a clear price on generative video. Volcano Engine’s Seedance 2.0 lists two token‑based tariffs: 28 CNY per million tokens when a video input is provided (video editing), and 46 CNY per million tokens for pure video generation without an input. The company says producing a 15‑second clip consumes about 308,880 tokens, which the vendor and some outlets round to roughly 15 CNY per clip — or about 1 CNY per second — under the higher rate.
The two price points reflect different technical loads: supplying an existing clip to be edited is billed at the lower rate because it reuses visual data and requires less fresh synthesis, while generating content from scratch consumes more compute and therefore costs more. Using the figures above, a 15‑second job would cost roughly 14.2 CNY at 46 CNY/million tokens and about 8.6 CNY at 28 CNY/million — small sums that nevertheless mark a step change from traditional video production economics.
The broader significance is twofold. First, Seedance’s pricing formalises a per‑token, per‑second market for video the way language models created one for text: it makes short, highly produced clips cheaply and predictably billable, opening the door for rapid experimentation by creators, marketers and small studios. Second, it crystallises the infrastructure economics of generative video: costs are not only monetary but also measured in GPU hours, queue times and downstream moderation workloads. Reports of long generation queues for Seedance indicate demand is already straining capacity.
For commercial users, the implications are immediate. Advertisers, social‑media creators and education platforms could amortise large creative campaigns at a fraction of legacy costs, changing agency workflows and pricing models. For incumbents in post‑production, the pressure will be to add value in strategy, curation and rights management rather than basic assembly. Meanwhile, companies that operate the GPUs and data centres that underpin these services will face incentives to expand capacity and optimise models for cost‑efficiency.
Risks and policy questions follow. Low unit prices accelerate the volume of synthetic video in circulation, compounding existing concerns about deepfakes, misinformation, copyright infringement and the provenance of training data. Regulators and platforms will have to balance innovation and access with scalable content moderation, clearer licensing regimes and technical provenance tools such as watermarking and metadata standards.
Seedance 2.0’s price announcement is not an endpoint but a marker of an industry in motion. The next phases will test whether lower per‑second costs translate into sustainable business models for providers and whether policy and product controls can keep pace with rapidly expanding creative capacity. For global markets, the move suggests that Chinese cloud and app ecosystems will continue to be influential suppliers of generative AI multimedia tools, shaping both competition and the rules of the road for synthetic content.