BrainCo AI Showcases Multimodal Agents at WAIC, Reshaping the Future of Creative Production

At the 2025 World Artificial Intelligence Conference (WAIC), Yao Ting, co-founder and CTO of HiDream.ai, took center stage to deliver a keynote address. He unveiled HiDream.ai’s advancements in multimodal agent technology for content creation, detailing both the technical breakthroughs and the burgeoning commercial applications. As an AI innovator focused on multimodal generation, HiDream.ai aims to transform content creation, shifting the focus from mere efficiency gains to a true revolution in productivity by “returning creation to inspiration and dedicating time to stories.”

The rapid advancement of AI technology is moving from labs to real-world applications at an accelerating pace. HiDream.ai is prioritizing the practical and is pioneering a path in commercial deployment based on “technology foundation, scene breakthrough, and value closed-loop.” The company believes that the true commercialization of AI isn’t about showcasing isolated technologies but empowering the entire chain, from model capabilities to service delivery and tangible results.

HiDream.ai is committed to a product development approach that emphasizes translating technology into value, and has built a progressive commercialization system of “MaaS-SaaS-RaaS.”

MaaS (Model as a Service) is the foundational bedrock. It involves building a multimodal foundation model with billions of parameters, which enables the generation and understanding of various modalities, including images, video, audio, and text.

SaaS (Software as a Service) acts as the bridge. HiDream.ai leverages its base models to create vertical-specific products and build creator platforms and communities, transforming technical capabilities into readily available services and lowering the barriers to entry for content creators.

RaaS (Result as a Service) is the ultimate culmination, offering services like commercial video marketing and new media creation agents that directly deliver “actionable results” for clients. This transforms AI from a mere “technical concept” into a genuine “productivity tool” for content creation.

This model – where services are backed by robust models and implemented within specific scenarios – has proven effective in real-world applications. The HiDream multimodal generation platform supports domains spanning film production, product marketing, and culture & tourism, creating a closed loop that links technology development and commercial value.

Multimodal Technical Leap: From “Capable of Generating” to “Generating Excellence”

Technically, HiDream.ai’s multimodal foundation model has gone through three major iterations, establishing core advantages of “deep understanding, precise control, and high image quality”. The model evolution spans from Version 1.0 in August 2023 (Diffusion Transformer, achieving multi-modal alignment), to Version 2.0 in June 2024 (Diffusion Autoregressive Transformer, enhanced time-space modeling), and finally to Version 3.0 in December 2024 (Mixture of Experts, multi-scene learning with memory enhancement). This trajectory reflects the continuous push to break existing barriers in generation technology.

These evolving capabilities translate into three core values: semantic consistency (ensuring consistent stylistic elements when bringing IP stories alive), precise controllability (supporting personalized customization and flexible element adjustment), and cinema-grade visual quality (4K resolution and stable long-sequence output). These advancements provide a robust technology platform specifically tailored for professional content creation.

In video generation, the model supports text-to-video, image-to-video, and start-to-end frame generation. It can accurately replicate styles, like that of domestic anime or Ghibli films by jointly learning camera and scene movements. By using the Diffusion Autoregressive Transformer (DiT+AR), HiDream.ai has addressed the “spatio-temporal consistency” challenge in video generation, meaning the generated content more closely aligns with the laws of the physical world.

Product Form: Agent-Driven “Creative Revolution” Reshapes Content Creation Workflow

In terms of product form, HiDream.ai centers on “intelligent agents,” building a toolchain that extends across image generation, video creation, and marketing communications.

Vivago agent, a short video second creation agent, emphasizes “multimodal input, intelligent decomposition, interactive generation” as a core advantage. A user simply provides images, videos, audio, text, etc. (e.g., a coffee shop logo, photos, promotional copy), and the Vivago agent automatically analyzes the request, decomposes it into tasks (storyboard design, script generation, asset retrieval), calls up the proper image/video generation model to supplement content, and integrates the output through an intelligent editing tool. It will not only understand the visuals of “a brown line-drawn flame + a wave logo,” but also capture the atmosphere of “a quiet, luxurious bar scene,” taking short video creation from “starting from scratch” to “on-demand generation.”

The product implementation of these multimodal capabilities has accomplished complementary functionalities in content creation: Vivago agent, focusing on short video recreation, enables users to quickly create individualized content by integrating template retrieval, intelligent editing functionality, and multi-modal generation. This helps overcome the problem of homogeneity that often arises with traditional template-driven creation. HiClip, on the other hand, tackles the problem of “content overload and inefficient distribution” in long-form video. It uses multimodal semantic understanding to extract core data from long-form video, realize highlight reel extraction, cross-platform editing capabilities, and inspire long-form video secondary distribution value.

AI’s value lies in connecting and empowering. The implementation of technologies and products requires strong ecosystem collaboration. Currently, HiDream.ai is working with partners in cross-border e-commerce, internet, film and television, new media, culture and tourism and other fields to build an ecosystem that encompasses multiple fields, thereby creating a “technology-scene-ecosystem” win-win situation.

HiDream.ai remains committed to empowering every creator to unleash their creative potential. By truly “understanding and assisting with creation,” AI is accelerating a revolution in content industry productivity. HiDream.ai anticipates that multimodal intelligent agents will serve as a pivot, exploring the potential of “technology as the pen and creativity as the ink” with industry partners, allowing every creator to focus on inspiration and allowing every story to reach far beyond.