Do I still need to make proxies?

Only when the editing machine cannot decode the camera codec in real time. The handoff itself works from the camera originals, because the editor reads the frames they scrub.

Will the media go offline and need relinking?

The mount is a normal path, so the paths in the project stay valid as long as both sides mount at the same place. There is no proxy set to attach and detach, so the usual relink step for a handoff does not arise.

Can two editors work on the same media at once?

Yes for media, which is read mostly. Project files are the careful case. Keep the project database local to each cutting room and exchange timelines as XML, AAF, or EDL, in the same way a shared storage workflow does.

Is this good enough for real time 4K playback?

That is not the claim. Sustained multi stream 4K over a thin link needs local media. The claim is scrub, review, conform, and pull the shots that make the cut.

Send the cut, not the drive

Mount the camera originals. Skip the proxy round trip.

By Marius-Florin Cristian · 18 August 2026

Your editor or colorist mounts the rushes where they already sit. Scrubbing pulls the frames under the playhead. The wire carries the shots someone opened.

No transcode before the handoff. No relink after it. No drive in a courier bag.

Built for a remote handoff

Scrub camera originals

BRAW, R3D, ARRIRAW, ProRes. Open them from the mount in Resolve, Premiere or Final Cut.

No proxy round trip

Ship the originals as they are, and take the timeline back as it is.

Only the selects move

Wire cost follows what got watched instead of what got shot.

Renders come back the same way

Write the export into the mounted folder. It lands on the other machine.

Small changes ship small

A 2 MB change inside a 1 GB file moved 18.56 MB, verified byte exact.

No cloud in the middle

Peer to peer, end to end encrypted. No per gigabyte egress bill.

Review without the upload

The approver opens the cut from the mount and watches it. No upload, no download, no link that expires.

Both sides can edit

Saves cross in either direction. A losing edit is kept beside the file as a conflict copy.

The handoff, priced

One hour of 4K RAW is 500 GB to 1 TB of camera negative. Its proxy set is 3 to 4 GB. Reaching that 3 to 4 GB is what takes the day.

Step	Without KEIBIDROP	With KEIBIDROP
Transcode to 1080p ProRes 422 Proxy	Hours of machine time before anything ships	Not needed
Upload the proxies	Per gigabyte, or an overnight upload	Not needed
Cut the material	Against proxies, until someone checks focus	Against the camera originals
Relink to originals	Path mismatches, media offline, reconform	Not needed
Or courier a shuttle drive	A day each way, one copy in transit	Not needed
Send the cut for approval	Upload, then a download at the other end	The approver opens it from the mount
Return the render	A second transfer service	Write it into the same folder

Four of those steps stop existing. The ones that remain move the frames someone watched.

The approval round trip

A cut goes to a producer or a client for sign-off, comes back with notes, and goes round again. Today each leg is an upload and a download of the whole file, and an 11 GB export costs half an hour in each direction before anyone can watch it.

Over a mount the approver opens the export in place. The bytes they watch are the bytes that move. Notes come back as a file written into the same folder.

Saves that cross both ways

Both machines can write. Measured over loopback on macOS with the collaborative profile on:

Save pattern	Convergence
Cold first read, 512 KiB	30 ms
Write temp then rename over the original, the usual application save	20 to 550 ms
Autosave storm, three renames back to back	80 to 540 ms, latest content wins, nothing lost
The other side edits your file	0.55 s
ffmpeg render on one machine, cold remux on the other, result back	about 1 s

Through that suite so far: vim, sqlite3, LibreOffice, ffmpeg, GIMP and pijul. Resolve, Premiere, Blender and Audacity open from the mount and are covered by the recording sessions, and the matrix run for them is pending.

Measured

Player paced at 3 MB/s, reading cold across a link at about 200 ms round trip, Windows client.

File	Read-ahead off	Read-ahead on	Trials
100 MB	5 stalls, 2.79 s frozen	0 stalls	n=10
1 GB	62 stalls, 41.35 s frozen	0 stalls, worst read 70 ms	n=5

A stream paced at Blu-ray rate, 40 Mbps, ran through nine encryption key rotations with zero underruns and a slowest fetch of 171 ms against a 2000 ms buffer.

Read-ahead removes stalls in sequential play. The first jump into a cold region still costs: a random seek up to about 2.8 s, and a large file whose header sits at the end 1.9 to 2.2 s.

Limits

Built for scrub and review. Scrub, review, conform, and pull the shots that make the cut. Sustained multi stream 4K for finishing wants local storage.
Proxies still fix decode. Make them when the workstation cannot decode the camera codec in real time. What goes away is making them to move the material.
Project databases stay local. Media shares well. A Resolve library or a Premiere Productions folder is a database, so keep it in the room and exchange XML, AAF or EDL.
Resolve and Premiere figures are in progress. The numbers above come from a paced player and a synthetic stream. Application measurements appear here with their conditions once recorded.

Common questions

Do I still need proxies?

Only when the machine cannot decode the codec in real time. The handoff itself works from the originals.

Will media go offline and need relinking?

The mount is a normal path, so paths stay valid when both sides mount at the same location.

Can two people work on the same media?

Yes for media. Keep the project database local to each room.

What happens if the connection drops?

Reads resume at the chunk they reached.

Video and audio projects between two computers · Run ffmpeg on remote files · How the read-ahead works