Radar Telegram alerts Open Radar

Another director: OpenDirector v0.1.0 based on vLLM-Omni and Minimax H3 ThinkFlowLab announced the first public release of OpenDirector v0.1.0 - a fully native platform for production, visual design and editing of short videos with integration of native heating engines. OpenDirector combines four stages: the user designs scenes in dialogue with the artificial intelligence assistant, structures references on an infinite canvas, creates storyboards and manages the timeline for the final rendering. The project is saved to the local disk, and media creation is done using local services or connected APIs. For local inference, the platform uses a combination of vLLM-Omni and MiniMax H3 (FastH3) models. Special built-in presets: three-view character creation, faces and objects to maintain consistency, as well as automatic assembly of 4-frame storyboards with multi-camera perspective. The editing block includes a non-linear timeline with clipping clips, adjusting playback speed and mixing audio tracks. Not only can the finished project be viewed in MP4, but also exported as timelines for FCPXML editing (the final cut is live), structured JSON or complete asset packs for development in a professional NLE. In summary: full local cycle: from text script and storyboard on canvas to non-linear editing and timeline export.

תרגם לעברית

Метаверсище и ИИще · · Source

Another director: OpenDirector v0.1.0 based on vLLM-Omni and Minimax H3 ThinkFlowLab announced the first public release of OpenDirector v0.1.0 - a fully native platform for production, visual design and editing of short videos with integration of native heating engines. OpenDirector combines four stages: the user designs scenes in dialogue with the artificial intelligence assistant, structures references on an infinite canvas, creates storyboards and manages the timeline for the final rendering. The project is saved to the local disk, and media creation is done using local services or connected APIs. For local inference, the platform uses a combination of vLLM-Omni and MiniMax H3 (FastH3) models. Special built-in presets: three-view character creation, faces and objects to maintain consistency, as well as automatic assembly of 4-frame storyboards with multi-camera perspective. The editing block includes a non-linear timeline with clipping clips, adjusting playback speed and mixing audio tracks. Not only can the finished project be viewed in MP4, but also exported as timelines for FCPXML editing (the final cut is live), structured JSON or complete asset packs for development in a professional NLE. In summary: full local cycle: from text script and storyboard on canvas to non-linear editing and timeline export.

Another director: OpenDirector v0.1.0 based on vLLM-Omni and Minimax H3 ThinkFlowLab announced the first public release of OpenDirector v0.1.0 - a fully native platform for production, visual design and editing of short videos with integration of native heating engines. OpenDirector combines four stages: the user designs scenes in dialogue with the artificial intelligence assistant, structures references on an infinite canvas, creates storyboards and manages the timeline for the final rendering. The project is saved to the local disk, and media creation is done using local services or connected APIs. For local inference, the platform uses a combination of vLLM-Omni and MiniMax H3 (FastH3) models. Special built-in presets: three-view character creation, faces and objects to maintain consistency, as well as automatic assembly of 4-frame storyboards with multi-camera perspective. The editing block includes a non-linear timeline with clipping clips, adjusting playback speed and mixing audio tracks. Not only can the finished project be viewed in MP4, but also exported as timelines for FCPXML editing (the final cut is live), structured JSON or complete asset packs for development in a professional NLE. In summary: full local cycle: from text script and storyboard on canvas to non-linear editing and timeline export.

A combination of local inference based on vLLM-Omni and the MiniMax H3 (FastH3) video model.

Consistency Presets: 3-view sheets of characters, objects, and multi-camera storyboards.

Professional export: Direct processing to MP4, export to FCPXML for editing programs and JSON OpenDirector repository on GitHub.

More on Radar

Want to see every report in real time?

Hundreds of sources worldwide, translated into your language, with alerts for what matters to you.

Sign up free
Share on WhatsApp Get alerts on Telegram