Run ffmpeg, Premiere, or any program on files from another computer
Your colleague has a 20 GB video file on their workstation. You need to run ffmpeg on it to extract a clip, check the codec, or transcode it. Or you want to open it in DaVinci Resolve, Premiere, or Blender. Normally you would wait for them to upload it, then download the entire file before you can start.
With KEIBIDROP, you connect to their machine and their files appear as a folder on yours. You run your tools directly on those files. The data streams from their computer as your program reads it.
Here is ffmpeg reading a video that lives on the other machine
$ ls ~/KeibiDrop/Mount/
raw-footage.mov audio-mix.wav project.prproj
$ ffmpeg -i ~/KeibiDrop/Mount/raw-footage.mov -ss 00:05:00 -t 00:00:30 -c copy clip.mp4
# ffmpeg reads from the virtual folder, streams the data from the peer
# Only the bytes ffmpeg actually needs get transferred
$ ffprobe ~/KeibiDrop/Mount/raw-footage.mov
# Shows codec info, duration, bitrate without downloading the whole fileffmpeg reads the file sequentially or jumps to specific byte offsets depending on the operation. KEIBIDROP serves each read request by streaming the corresponding bytes from the peer. For an operation like ffprobe or extracting a short clip, only a small fraction of the file actually transfers.
Every program that reads files can use the folder
The virtual folder works at the filesystem level, so your operating system sees an ordinary directory. Every program that reads files can use it:
ffmpeg, ffprobe, VLC, QuickTime, DaVinci Resolve, Adobe Premiere, After Effects, Blender, Photoshop, GIMP, VS Code, vim, git, Python scripts, shell pipelines. If the program can open a file from a folder, it works with KEIBIDROP.
Reading a header off a 50 GB file moves a few megabytes
If you run ffprobe on a 50 GB file, only a few megabytes transfer (the header and metadata). If you extract a 30-second clip from the middle, only the bytes for that segment transfer. If you open the file in a video editor and scrub through the timeline, each frame loads on demand.
This is different from syncing the whole file first. You start working immediately and the transfer footprint matches what you actually accessed.
You can write the output back into the same folder
You can also write the output back. Run ffmpeg to transcode a file and save the result to the shared folder. It appears on the peer's machine. Both sides can share files and results at the same time.
The connection goes straight between the two machines
The connection goes directly between the two machines, encrypted end-to-end. The encryption is quantum-resistant, no cloud storage is involved, and the files never leave the two devices (unless a relay forwards the encrypted stream through firewalls).