a distributed video processing tool, based on FFmpeg and VapourSynth
We cut the original video into multiple clips, and then process each clip in parallel in queue order. After all clips are processed, we merge them into the final video.
- docker and docker-compose
- Nvidia GPU / AMD GPU with ROCm support (7000 series on WSL2)
- GPU container runtime (optional)
- make (optional)
edit Line 19 to your LAN IP address
docker-compose -f deploy/docker-compose/lite/docker-compose.yml up -d
It will run all containers in a single host, then open http://localhost:8989
in your browser to access the dashboard, open http://localhost:8080
to access the Asynq monitor.
We use this container as the base image, which contains FFmpeg, VapourSynth, PyTorch...
So in the dashboard, we can select a compatible script to process the video!
Deploy the system in a distributed way, you can refer to the docker-compose directory for more details.
first, run docker-compose-base.yml to start the basic services, and open Consul dashboard, add a K/V pair with key finalrip.yml
and value is the content of finalrip.yml -- or your own configuration file
then, run docker-compose-server.yml to start the dashboard, server, cut worker, and merge worker services
finally, run docker-compose-encode.yml to start the encode worker services, we can deploy multiple encode workers in different hosts to speed up the encoding process
Note: we suggest that deploy oss service, cut & merge worker in the same host
Override the default configuration by setting:
Special Env Variables:
FINALRIP_REMOTE_CONFIG_HOST
Consul host, default is None, that means load config from local file. When set, it will load the config from the Consul K/V store. When set toEASYTIER
/easytier
, will try load config fromFINALRIP_EASYTIER_HOST:8500
(10.126.126.251:8500
by default).FINALRIP_REMOTE_CONFIG_KEY
Consul key, default isfinalrip.yml
In vs-playground, we provide the same environment as the encode worker, so you can develop and test your script in the playground.
- API document: here
- Build your own
encode
image: refer to the vs-playground, and set theTemplate Repo
name in dashboard if you wanna select a script from the repo'stemplates
folder.
make all
make pt
make pt-rocm
for AMD GPU
This project is licensed under the GPL-3.0 license - see the LICENSE file for details.