LibTV Enters Top 3 in AI Creation Tool Traffic, Agent Feature Launches to Lower Video Creation Barriers

According to June data for the AI creation track, LibTV ranked among the top three web tools with approximately 1.9 million monthly visits, becoming the only product in the top 10 to achieve month-over-month growth. Meanwhile, LibTV officially launched its video Agent feature, supporting natural language-driven end-to-end video generation and incorporating over 100 professional Skills, aiming to lower the barrier to video creation.

Traffic Data: LibTV Bucks the Trend

• According to Qubit reports, in terms of June web monthly unique visitors, Jimeng AI led with about 1.5 million, but declined approximately 25% month-over-month, marking four consecutive months of decline.
• In monthly visits, Jimeng AI held a commanding lead with about 9 million, Gaoding AI ranked second with about 2 million, and LibTV entered the top three with about 1.9 million, growing approximately 8% month-over-month—the only product in the top 10 to see growth.
• Multiple products such as Gaoding AI, Kling AI, LiblibAI, and Tencent Hunyuan saw unique visitor declines exceeding 20%. Yingmo Technology Hyper3D surged over 100% month-over-month, becoming one of the few growth drivers.

Agent Feature: From 'One-Sentence Generation' to 'Deliverable Videos'

• LibTV launched in March and introduced the Agent feature three months later. The product lead stated that the team waited for user feedback to clarify the product direction.
• The Agent allows users to select Skills via natural language, automatically completing the entire workflow: creative planning, storyboarding, asset generation, editing, music scoring, subtitling, etc., outputting a video ready for publication.
• The final video must meet structural completeness (intro, outro, subtitles, music, audio-video sync) and narrative completeness (viewers can understand the theme and emotion).
• The generation process adopts a Human-in-the-loop design: the Agent first outputs narration script and storyboard plan; after user confirmation, assets are generated. For modifications, users can pinpoint specific storyboard segments for rework without restarting the entire process.

Skill Ecosystem: 100+ Professional Templates Lower the Barrier

• LibTV has accumulated over 100 Skills covering professional film and TV, commercial advertising, short dramas and comics, music videos, etc., and is called 'the world's largest professional video Skill Hub.'
• Skills fall into two categories: filling the lower limit (solving basic scenarios where models perform poorly) and raising the upper limit (combining mainstream aesthetics to form distinctiveness).
• Users can create custom Skills, encapsulating personal experience into reusable workflows. Skills are deeply integrated with the Agent framework, making them difficult to replicate simply.

Product Architecture: Dual Views and Multi-Agent Drive

• LibTV Agent supports dual-view switching between Storyboard and Node workflow, balancing creative efficiency and fine-grained control.
• After generation, users can make local adjustments based on the storyboard and timeline, such as regenerating frames, modifying subtitles, or adjusting order, without starting from scratch.
• Multi-Agent driven, it can handle various inputs including text, images, video, audio, and documents, automatically decomposing tasks into scripts, shots, and frames.

Impact and Outlook

• The Agent feature aims to serve creators who 'have ideas but lack professional skills,' lowering the threshold for video production.
• The product lead believes that the Skill ecosystem in the multimodal domain will form a long-term barrier because there is no single answer to aesthetics—the richer the styles, the more demands can be addressed.
• User data will be accumulated as preferences and Memory, enabling the Agent to better understand individual aesthetics, forming a continuously growing creative experience network.