Check out the Docs! ✨ https://in03.github.io/proxima/
Launch the worker on as many computers as you have free and it'll pool all the free CPU threads together to encode multiple proxies at once. Only have the one computer? Encoding runs entirely on the CPU, leaving GPU-heavy Resolve with plenty of resources to continue editing while you pump out proxies. Once they're finished, they're automatically linked.
DaVinci Resolve's greatly benefits from having all-intra media, like ProRes or DNxHD. If you shoot in h.264 or h.265 like many do, you're likely to see great performance improvements using proxies. This application makes queuing, encoding and linking them quick and easy.
I started this for the company I work for, well before BPG was on the scene. If BPG works better for you, go for it! As it stands BPG won't do any all-intra codecs on Windows, which is a dealbreaker for us. It also works on a watch-folder basis with no filename whitelisting or path-filtering rules. That means EVERY video file becomes a proxy, whether you need it or not. Our workflow often sees the shooter doing a rough assembly of chosen takes as an exported timeline. We simply import this timeline and queue proxies from it. If you work with chronic-overshooters, you'll save a heap of disk space and encoding time queuing proxies from a roughly-organised timeline.
Note
Only Davinci Resolve 18 is supported
Please check out the "resolve-17" branch if you need Resolve 17 support. Prior versions are untested, but may work depending on Resolve's API feature set for that version.
Proxima is composed of three major parts:
The 'queuer' and 'worker' are bundled together in the CLI app. They are both installed from the same source, called from the same command and share the same configuration file. As such, any computer that has the CLI app installed can both queue proxies (so long as Resolve is set-up) and run workers. Install it with pipx:
pipx install git+https://github.com/in03/proxima
The broker is best installed on an always-on computer or server. If it's not running neither queuers nor workers can communicate. Very little configuration is required. Just make sure it's accessible over LAN. Redis and RabbitMQ are Celery's best supported broker backends. Only Redis has been tested extensively with Proxima.
Install Redis with docker:
docker run -d --name some-redis -p 6379:6379 redis-server --append-only yes
An in-house web-app monitor for Proxima jobs is coming and will include some really helpful, specific features. Until it's ready, if you want to monitor your jobs, it's a good idea to install Flower - an official monitor for Celery tasks and workers. Install it with docker alongside your broker:
docker run --name flower -e $CELERY_BROKER_URL=redis://192.168.1.171:6379/0 -e FLOWER_PURGE_OFFLINE_WORKERS=300 -d flower
Set the IP address and port to the same as your broker container. Don't forget the Redis protocol redis://
.
Consider setting FLOWER_PURGE_OFFLINE_WORKERS
if you don't have a well-defined set of computers to run the workers, or there's potential to change worker hostnames. 300
for example, would clear the list of offline workers every 5 minutes. This just keeps things tidy.
Usage: proxima [OPTIONS] COMMAND [ARGS]...
Options:
--install-completion Install completion for the current shell.
--show-completion Show completion for the current shell, to copy it or
customize the installation.
--help Show this message and exit.
Commands:
link Manually link proxies from directory to source media in open...
mon Launch Flower Celery monitor in default browser new window
purge Purge all proxy jobs from all queues
queue Queue proxies from the currently open DaVinci Resolve timeline
work Prompt to start Celery workers on local machine
celery Inject Celery commands to control tasks and workers
On first run, you'll be prompted to alter your settings. The app will copy the default settings to the OS user configuration folder.
$XDG_HOME_CONFIG/proxima/user_settings.yml
(may not open settings automatically)%homepath%/proxima/user_settings.yml
proxy_path_root
All proxies will be encoded to this directory. They'll retain the source media's directory structure:
paths:
proxy_path_root: R:/ProxyMedia
celery
Celery runs all queuer/worker/broker communications. Make sure you set all of the below addresses as per your environment!
broker:
url: redis://192.168.1.19:6379/0
job_expires: 3600 # 1 hour (cleared if not received by worker)
result_expires: 86400 # 1 day (Needed for webapp monitor)
Warning
Make sure you set
result_expires!
to a reasonable value otherwise Redis may run out of memory! If you need persistent results, consider configuring Redis for persistence.
worker
worker:
loglevel: INFO
terminal_args: [] # use alternate shell? Recommend windows terminal ("wt") on Windows.
celery_args: [-l, INFO, -P, solo, --without-mingle, --without-gossip]
Queuer and worker have separate loglevels to make debugging a little easier if you've got just one worker playing up.
Some pretty dangerous concurrency settings were moved out of the configuration settings to prevent accidents. Windows doesn't support pre-forking for concurrency. Actually, Celery doesn't officially support Windows anymore at all. Mac and Linux can pre-fork, but the worker launcher invoked with the proxima work
command gets around this by launching separate processes with workers named worker1@host
, worker2@host
, etc. We're going for the lowest common denominator here. This works fine with Mac and Linux too. It makes monitoring easier and cross-platform behaviour more consistent. For those concerned about the extra overhead who would like to squeeze out every last bit of performance on Mac or Linux, consider injecting the celery multi
command with proxima celery
. See Celery daemonization
Hey! Thanks! Any help is appreciated. Please check the Contribution Guide.