Inspiration

Families keep years of phone and camcorder video on their NAS, and none of it has subtitles. Hard-of-hearing relatives miss the words, grandparents and grandchildren who speak different languages can't follow each other, and silent clips say nothing about who, where or when. Cloud captioning would mean uploading private family footage. A NAS with on-device AI already has everything needed to fix this at home.

What it does

Ask the NAS assistant: "Subtitle Family/2024 in English and Spanish." For each video, Subtitle Everything picks the right source:

  • Downloaded: published films and episodes get a matching subtitle from the UGOS subtitle service.
  • Generated: home videos get subtitles created on the NAS from the speech.
  • Described: clips with no speech get a short bracketed cue from the AI Album index, such as [Beach · July 2024 · sunset, waves].
  • Translated: every other requested language, cue by cue, with the original timing.

It shows a preview table, and only after approval writes standard .srt files next to each video. Videos and existing subtitles are never changed, and running it again only handles new videos.

How we built it

The Skill is pure orchestration written to the Uliya Skill Development Standard. It delegates each step to UGREEN's registered experts: file-agent for the inventory and writing files, media-agent for title lookup and subtitle download, knowledge-agent for transcription and the run log, photo-agent for silent-clip labels, and document-agent for translation. Every state an expert can return (completed, failed, cancelled, waiting for the user, waiting for approval) is handled, and it passes the kit's static checker with no errors or warnings.

Challenges

Making it safe by default: subtitles must never overwrite an existing file, translations must keep exactly the same cue count and timing, and nothing may be invented when a clip has no speech or the transcript has no timestamps.

What's next

Validating on a real iDX6011 Pro: transcription of video files with segment timestamps, subtitle download, and Spanish on-device quality. After that, timed scene-by-scene descriptions for silent clips.

Built With

  • machine-translation
  • markdown
  • nas
  • speech-recognition
  • ugos
  • ugreen
Share this project:

Updates

Submission history