From Creative Intent to Finished Video: How QuVideo Brings Video Creation Agents to Smart Devices

After a trip, a phone may contain hundreds of photos and videos. Turning them into a short video with a coherent story, however, can be a daunting task for many users.

How can AI take over this process? At QCon Software Development Conference 2026 Shanghai, Li Linlin, VP of Business at QuVideo, shared the company's approach to bringing video creation agents to smart devices in a talk titled Scaling Video Creation Agents for Hardware Scenarios.

QuVideo's solution allows users to simply describe what they want, such as "Turn my photos and videos from last week's trip into a three-minute story-driven vlog." The agent then selects relevant footage from the local album, plans the story and generates the finished video.

This is the latest development in QuVideo's Intelligent Video Creation Solutions, designed to help consumer electronics and imaging device manufacturers extend their video creation capabilities while making personalized video creation easier for everyday users.

Edge-Cloud Collaboration for Different Devices

Video creation involves multiple stages, including content understanding, story planning and rendering. Relying entirely on the cloud increases data transfer and processing costs and raises privacy concerns. Relying solely on the edge, meanwhile, is constrained by computing power, memory and power consumption.

To address these challenges, QuVideo adopts an Edge-Cloud Architecture that divides the process into three stages: Edge Content Understanding, Cloud-Based Creative Planning and Edge Rendering.

First, visual recognition algorithms running at the edge analyze local photos and videos, identify relevant materials and extract key moments to build structured content context. The cloud-based agent then combines this information with the user's intent to plan the story and generate a creative configuration. Finally, the device's video engine interprets the configuration, builds the timeline and handles editing and rendering to produce the finished video.

This architecture assigns frequent local processing tasks to the edge, while the cloud handles more complex intent understanding and creative decisions. The system can also adjust how tasks are distributed based on device performance and network conditions, supporting a range of devices, from flagship to mid-range models, while balancing user experience and operating costs.

Turning 14 Years of Video Creation Experience into Scenario-Specific Skills

While the Edge-Cloud Architecture defines where tasks are processed, the quality of the final video depends largely on how the Agent plans the story.

Traditional one-click video creation typically relies on predefined templates and follows relatively fixed workflows. QuVideo, by contrast, has distilled its 14 years of video creation experience into scenario-specific skills. These skills translate professional creative knowledge into reusable rules, defining the inputs, constraints and outputs for different scenarios. The agent can then select the most suitable skill based on the user's intent and available materials.

Travel videos are one example. An itinerary provides a basic storyline, but actual footage may be unevenly distributed, and important interactions between people may not take place at major stops. Skills guide the agent in identifying which moments to keep, adjusting the story when footage is limited, and connecting different clips into a coherent narrative rather than simply stitching them together. Music selection also takes the emotional flow of the story into account.

These rules help preserve important people and moments, create a clearer storyline and better match the soundtrack to the user's expectations, while reducing creation time and the need for revisions. User feedback also provides a basis for continuously improving the skills.

Reliable delivery is equally important. A unified workflow coordinates each stage, from content understanding and story planning to video generation, ensuring smooth execution between the edge and the cloud. If a task times out or is interrupted, the system can automatically retry and recover.

For consumers, video creation agents offer a simpler creative experience and more personalized, story-driven videos. For hardware manufacturers, the same capabilities can support multiple scenarios and device models, creating new opportunities to improve software experiences. Delivering these benefits requires more than models alone: it takes the combined capabilities of content understanding, professional video creation knowledge and a reliable engineering system.