Dreamina is launching its agent Octo into closed beta.

ByteDance is launching Octo, a Dreamina Agent and Seedance 2.0 video model overlay, in a closed beta for creators with applications closing on May 15.

ByteDance and CapCut have posted the closed beta application form for Octo (also referred to as Dreamina Agent on the website), a conversational agent integrated into the Dreamina platform and designed as an overlay to the Seedance 2.0 video model. Selection is application-based, with applications closing on May 15th and activation of selected accesses scheduled by June 1st.

Its approach differs from traditional direct prompting: the user interacts with Octo in a chat panel, provides it with references (images, storyboards, mood), and delegates the planning, generation, and refinement of outputs to it. The agent adapts in real time to changes in direction, maintains the context of previous iterations, and progressively learns the user's stylistic habits. For images, it can produce up to 40 visuals in a single sequence while maintaining overall coherence.

Regarding the underlying model, Seedance 2.0 takes text as input, as well as up to 9 images, 3 short videos, and 3 audio files, and generates 1080p video with attention to character and IP continuity, style transfer, visual effects, lip-sync, and narrative coherence over time. The engine remains separately accessible for a free trial on the Dreamina website.

Full access to Octo's features will be free during the beta. Dreamina indicates that the product will transition to a paid model upon official launch, with pricing to be announced beforehand. The form asks applicants about their role (animation, indie game development, web fiction, brand design, social creators, advertising, post-production), their distribution channels, subscriber counts, and portfolios, which suggests a selection geared towards established or active creators.