ClipLogger is a storage agnostic Digital Asset Manager. Construct a custom metadata schema, use local or cloud vLM models to suggest tags, then commit them forever to a file name.
Grid thumbnail for scanning, Reel for reading a whole shoot as filmstrips, and Logging for going clip by clip. Same metadata underneath — pick whichever matches what you are actually doing.
Send a batch to our cloud when you want it faster. One credit, one clip, priced before you press go. The same job runs free on your own Mac if you’d rather wait.
Nothing you confirm is write-only. Every value becomes a way back — what the model read, what you decided, who was in frame, which takes you flagged. Stack them and a 400-clip card turns into the eleven shots you were looking for.
Every clip becomes a filmstrip, in order, at a glance. Scrub any of them in place. Then send the selects out as FCPXML or EDL and finish wherever you already finish.
It reads the frame, offers a value for every field, and writes none of them until you say so. Recognised subjects are identified with a bounding box.
Every field carries a type, a definition in your words, its own options, and a declared mapping to a real metadata field. Import it, export it, hand it to an editor who has never opened ClipLogger.
Import a CSV and customise field concatenation. Hold shift to bring up the Code Replace search; enter commits the selection to the output field.
The camera gave you C0001. ClipLogger uses confirmed metadata to compose an output name from the fields you chose, in the order you chose them — then executes it to disk, renaming the files themselves so the description survives every app you hand them to.
Content-based verification, an MHL manifest, tiered destinations, custom renaming, and preview derivatives generated in flight. ClipLogger recognises a card it has already taken and refuses to copy it twice.
ClipLogger keeps your metadata next to the clip, in an open standard. No lock-in to break out of, because there was never a lock. Delete the app tomorrow and every field you confirmed is still sitting beside your footage.
{
"content_key": "3f2a9c1d5e7b0846:25017860096",
"capture": { "date": "2026-07-16", "reel": "082" },
"log": {
"take": "flagged",
"action": ["touchdown", "gain", "celebration"],
"shot": "wide",
"location": "near-side",
"subjects": ["Tre'Quan Smith", "McKenzie Milton"]
},
"provenance": {
"engine": "local", "model": "qwen2.5-vl-7b",
"confirmed_by": "you"
}
}
This is not a feature comparison — it is a question about where the work you paid for lives afterwards. Both models cost money. Only one of them leaves you with something when you stop paying.
You pay us to go faster. You never pay us to get your own work back — there is no version of ClipLogger where that is the offer.
Not unless you choose Rush for a specific run. The default engine is on your Mac, and any run that spends credits shows you the clip count, the cost and your balance before it starts. If you keep footage on a JuiceMount server, ClipLogger also publishes what it derives back to that server for your team — one switch in Sharing & Privacy, on by default.
Entirely, once you are set up. Offload, browse, logging and playback never need a connection. The local AI models download once and then run on your Mac offline for good; Rush and account sign-in are the only parts that need to be online.
Yes — bring an OpenRouter key and the request goes from your Mac straight to the provider you picked. We never see it.
Yes, up to about fifteen minutes per clip. No per-minute maths and no surprise line items — the cost is shown before the run starts.
Nothing. They are files beside your footage, not rows in our database. The viewer, offload, browse and logging tiers stay free regardless.
Not during ingest. Your card is read to be copied, never written to: every copy is checksummed against the original and any renaming lands on the copy. There is also a separate Rename bench for files already on your own drives — that one renames in place by design, with a full preview before it runs.
The metadata format is open and documented, and your sidecars are plain files you can read without us. The application itself is not open source.
A Mac with Apple silicon on macOS 15 or later. There is no Intel build.
Ingest, browse, log and analyse on your own Mac — private by default, metadata in open sidecars you keep forever. Pay only when you want our cloud to do it faster.
macOS 15+ · Apple silicon · free tier needs no account