Twenty-five years of work leaves a lot behind: clients, projects, invoices, photos, and an enormous pile of video. Somewhere in there is the clip, the contact or the answer you need. The problem is finding it, and trusting what you found.
The idea I am working toward is a golden record: one authoritative, indexed answer to “what do I have, where is it, and is it the real one?”
Three pieces
- Fingerprinting. Every file gets a cryptographic hash. Identical files are exact duplicates, no matter what they are named or where they live. A changed file is a different file. That makes dedupe and verification provable instead of a guess.
- Indexing. Metadata is pulled from the files themselves: date, camera, location, duration, format. On top of that, local AI helps describe and tag what is in the footage, so a search for “aerial, church, autumn” works.
- Relationships. Files connect to the projects and clients they belong to, which is what the planned CRM is for. A clip is not just a clip. It belongs to a job, a person and a date.
Rules I hold to
- Originals are never modified. The index points at them. It does not replace them.
- Everything the AI tags is a suggestion with a confidence level. Anything important gets a human look.
- It all runs locally, on hardware I control. Client material stays out of the cloud unless a client asks for it.
Where it leads
Once the record exists, good things follow: safe cleanup, reliable backup, faster delivery to clients, and the ability to answer a question like “do we have footage of that building from three years ago?” in a minute instead of a weekend. It is also the foundation for the next step, compressing the archive safely.