Open-source Agent Skill faceless-shorts turns scripts into vertical videos

On 2026-11-12 at 10:04, the AI section of DEV Community reported that a developer has released faceless-shorts, an open-source Agent Skill released under the MIT license. The tool targets the production of faceless Shorts, meaning vertical videos that carry a voiceover, a headline and captions while no person appears on camera.

According to the report, faceless-shorts converts a short, scene-by-scene script into a finished 1080x1920 MP4. During the pipeline, every scene is voiced through the Miragea speech API, and the duration of each scene is determined by the actual length of the generated audio rather than a preset figure. Word-highlight captions are then burned directly into the video.

On the runtime side, the report states that only Python and ffmpeg are required. No GPU is needed, and the workflow does not depend on a stock-footage key or a second LLM key. Input is written as a structured script, with scenes defined through fields such as {"scenes": [ {"id": "hook", "keyword": "Hone, which keeps each segment organized. If a single line is revised, the tool re-records only that line rather than regenerating the whole video. No further developments, user adoption figures or follow-up announcements are covered in the available material, so the project is currently documented at this initial release stage.