Reelful's Agentic Video Editor Aims to Democratize Content Creation

Many individuals capture photos and videos at events, trips, and conferences but rarely share them online, a problem identified by Ekaterina Deyneka. This reluctance stems from the perceived difficulty, tediousness, and manual nature of video editing.

Kate highlights this "art-like" manual process as the core barrier preventing people from posting content they've recorded.

Agentic video editing simplifies the process: users input their media (photos and videos) and provide minimal context or directions. The AI agent then takes over, understanding the media's content and identifying the most relevant moments.

The agent assembles the video, generates captions, adds music, incorporates voiceovers, and includes B-roll footage, delivering a ready-to-share clip. This automation also handles tasks like removing pauses or less successful shots in videos where someone is speaking to the camera.

From an infrastructure perspective, agentic video editors share strong similarities with agentic app builders. Both systems rely on a 'sandbox,' a remote machine that spins up to provide an isolated environment for the agent.

Within this sandbox, an agent equipped with specific tools and skills executes the user's requested tasks. The main difference lies in the output: app builders produce code-based application previews, while video editors render completed video compositions.

Reelful's primary focus is on editing existing footage rather than generating AI content from scratch, a task that presents unique challenges. The agent must meticulously curate content, deciding which moments are best, what to omit, and how to structure everything cohesively.

Furthermore, real-world footage can often be messy, incomplete, or of varying quality. Despite these imperfections, the agent is tasked with transforming such raw material into a highly polished, professionally produced final video.

The Reelful workflow begins with users submitting their media along with a prompt or directions specifying how they want the video edited, with the goal of producing a polished clip. The initial step involves comprehensive media understanding, where the system identifies events and key elements within the submitted footage.

Following media understanding, for speaking-to-camera footage, the system performs transcription. Finally, it generates a creative plan, which is then presented to the user for approval before actual editing commences.

A critical part of the Reelful process is generating a detailed plan before any actual editing begins. This plan outlines the proposed structure and elements of the video.

Once the user approves this creative plan, a dedicated 'sandbox' environment—a remote machine—is spun up. This isolated environment provides the necessary resources and tools for the agent to execute all the planned editing tasks.

Reelful's agents are equipped with specialized skills crucial for video editing, including defined cut rules that guide the selection of the best moments from raw footage. They also possess knowledge about appropriate font pairings, discerning which fonts suit particular use cases and which do not.

Additionally, agents handle B-roll generation, a process where aesthetic judgment and craft are essential. Further sub-processes include the integration of music, voiceovers, and dynamic image animation to enhance the final video.

Reelful utilizes Remotion, an open-source framework that enables the creation of videos as React code. This approach treats a video as a code file, defining the order of assets, tracks, and their progression, making it highly structured and programmable.

AI agents are exceptionally good at writing code, which makes Remotion an ideal tool for them to construct and manage video compositions. A verification layer is integrated to ensure that these code-based compositions are free of errors before rendering.

Reelful adopted a mobile-first strategy to enable users to edit videos conveniently on the go, such as while driving, walking, or during other daily activities. Recognizing that complex prompting for videos can be challenging for many users, Reelful developed directional templates.

These templates simplify the process, allowing users to simply drop in their media without needing intricate prompts. A built-in editor further enhances the user experience, providing a familiar interface for making final adjustments and tweaks.

Reelful has released version 2.0, a new beta version, exclusively for conference attendees, encouraging them to try it and provide feedback. This launch follows significant funding from A16Z Speed Run, supporting the company's development and expansion.