PLAN / LOCAL TOOL
Video Tagger
Go through a whole folder of footage on your PC: transcribe every video, tag it and describe it, all saved in one library.json. It also writes library-for-ai.md: every video’s name, tags, description and full transcript in one file to give to ChatGPT or Claude.
Start Video Tagger
- On your Windows computer, open the repo’s
local-taggerfolder and double-click Start Video Tagger.cmd, or use the button below to copy the command. - Pick a folder of videos. Subfolders are included.
- Leave the window open while it works. It shows each video and step, then a summary.
Run it again any time: videos already in the library and unchanged are skipped, and a stopped run carries on where it left off.
What you get for each video
- DetailsLength, vertical / horizontal / square, resolution, frame rate and recorded date.
- TranscriptTXT, SRT and word-timing JSON from the Transcribe Video engine on your GPU.
- Speech levelNone, little or lots, so B-roll and trail footage are easy to find.
- TagsOnly from your tag list. Anything else becomes “other” with a suggested label.
- DescriptionOne or two plain sentences: what’s said, then what’s on screen.
- Needs reviewFlagged when the model is unsure, with the reason.
- One file for AINames, tags, descriptions and full transcripts of the whole library in
library-for-ai.md.
Your tags
Edit D:\VideoTagger\tags.json to change the list. The starting tags:
- trail review
- race
- training run
- long run
- workout
- strength
- gear
- food & fueling
- Hooper
- RRC / run club
- talking to camera
- B-roll
- scenery
- aid station / crew
- travel
- other
Your videos are only read, never changed or moved. Transcription uses the same engine as Transcribe Video, and tags come from a local Ollama model that looks at the transcript and a few frames. Nothing is uploaded and no paid API is used. Results go to D:\VideoTagger.