Barista - Home Cloud Rendering For Blender

by Dennis Brown in Addons


Item Rating

This item has an average rating of 4 from 21 ratings by the community.

58%
20%
5%
15%
5%
  • Winfried Weis
    over 1 year ago

    Thank you!

    This addon allows the user to render efficiently via AWS. So far I have not experienced any major issues, and the few problems and questions I have had have been fixed as quickly and competently as possible by the developer.

    By developing this addon, the community is given the opportunity to complete projects in AWS without much prior knowledge when their own computing power at home is insufficient or otherwise needed.

    I now also use the addon for client work and to complete my projects faster. For me worth full 5/5 stars and I would have paid a higher price to support the development further.

    Of course you have to deal with Amazon support to get more processing power. But that's not the developer's fault. For me one of the greatest addons there is. It makes my life easier every day.

  • Timur
    about 2 years ago

    I tried so many different Blender files (stills/ movies) and instance types. I tried this app on several Mac OS systems (Intel and Silicon) and with different Barista releases, but unfortunately the most of my render attempts don't work. I doesn't give me more then a progression bar what keeps staying on 0.00%, even if in the AWS Console is showing that the instances are running. I would love to have this app working, but it just doesn't! So said!

  • ARIF
    over 2 years ago

    Hey,

    I love this little tool. So useful. There seems to be an issue with OpenImageDenoiser in the latest Blender versions and Barista. If I enable it, Barista doesn't recognized it as an option and spits out the following.

    No module named 'AI_denoise_addon' ModuleNotFoundError: No module named 'AI_denoise_addon'

    The Linux 64 versions should have this option as well. However I did try to render without the openImagedenoiser and still got the error, looks like it doesnt render at all with the option.

    Do you have a git? or anything to follow these issues?

    Best,
    Arif

    • Dennis Brown

      over 2 years ago

      Sorry for the confusion Arif!

      The error you're receiving is from the local userpref.blend file referencing an older add-on. This has since been fixed in 1.7.3.

  • steve givens
    almost 3 years ago

    not to lie is really hard to use and may need help with using this cuz im running a gtx 660 2gb but is still slow

    • Dennis Brown

      over 2 years ago

      Hello Steve!

      Until the 1.7.1 update, many people were mistakenly attempting to render Eevee renders with CPU instances, which would show such slow speeds. Many features of Blender do require you to use a GPU to render at a good speed, and Barista was updated in 1.7.0+ to let you know when this will be an issue. Though this isn't the only cause, there can also be issues with the settings in the scene file, various cycles features, denoise, and anything else that may require additional configuration in the scene, or a GPU. Again, please reach out for suggestions to get the best render performance. That's what I'm here for 🙏

  • Jeff Lange
    almost 3 years ago

    This program has unlocked the god-like power of using AWS. It takes a minute to understand, but honestly, that's nothing compared to the years it would take to understand AWS on your own. Well worth the money!

  • regono
    over 3 years ago

    I'm loving the simplicity of setting up renders. It is so much easier than Brenda web workflow.

    I am having issue with a scene of mine. I hope you can help. It renders slowly on my GTX980 with only 4gb video memory so when I send it to G3.8xlarge I expected it to render faster. But instead I am getting an out of memory error. The hardware on this instance is everyway better than mine so what could be causing this? I tried increasing the EBS volume to give it more storage too and that didn't work.

    This is the error,

    Out of memory in cuLaunchKernel(cuRender, 1, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0) (device_cuda_impl.cpp:666)Blender Crashed

    • Dennis Brown

      about 3 years ago

      Sorry for the late reply!

      At one point the instances had issues with the CUDA support in the older 1.5.x version (now depreciated). Since then we have updated the support for the G3 and G4 instances specifically for better performance, but if this was an Eevee scene, Blender does not offer multiple GPUs. Because of such confusion, I've added more dialogs and messages in the app to help make sure these sort of issues can be avoided before the render.

      If you still have issues, please send a question and we can look to see what's happening, and what we can do to get around the issue for you. 👍

  • primaryfantasy
    about 4 years ago

    Hi everybody! Today I decided to replace my aging Brenda setup with Barista. Getting it up and running took the same amount of time as the tutorial video. The first render was an immediate success.
    So far I'm very happy with the GUI and how easy it makes rendering on AWS. I encountered two problems however, and also have some questions and comments.

    First the problems:

    During my test, Barista failed to upload 3 frames to the s3 bucket.
    As a test, using Blender 2.83 and Cycles, I rendered 20 frames to multilayer EXR, and PNG via a file output node in the compositor. Then I first started a single g3s.xlarge instance and waited for two frames to be completed, then stopped rendering, and restarted from frame 3, after requesting an additional 3 g3s.xlarge instances, so 4 instances in total. It proceeded to render all frames up to 20. When I checked, it had quit one of the instances, while 3 kept running with status "request canceled". PNG frames 1 to 20 were in the s3 bucket, for EXR however, only frames 1 to 17 were uploaded to the bucket. The 3 instances were still running half an hour later, so I stopped them manually. I didn't see any error messages.

    Another minor problem I observed was, that during my 20 frame test render, Barista stopped auto-downloading frames after a while (around frame 4 I think). I then used the green cloud icon to batch download them, but according to the file modification dates, it seems it also re-downloaded the ones it already auto-downloaded earlier. Not a big problem with a few frames, but it would be nice if it could detect already downloaded frames and avoid unnecessary file transfers.

    Some questions and suggestions:
    - How does Barista manage the render queue? Can I close the application and the queue will continue on AWS, like it did with Brenda?
    - Is it possible to add instances after rendering already started?
    - Is it possible to queue multiple cameras, scenes, or blend files at once?

    • CG Cookie

      over 3 years ago

      Thank you for your feedback!

      All of the download issues have been remedied with the new redesigned render system in 1.6.0: While we don't support multiple renders yet, our new render system does provide what we need to support in the immediate future. In the meantime, others will open multiple Barista clients for multiple projects.

      In the previous versions, you could only turn off the client during a render when using OnDemand instances. 1.6.0 allows you to do this same thing with the cheaper instance option (Spot Requests), which is now the default instance.

      You cannot still cannot add instances to running renders, but you can start your render with a small block of frames to gauge the speed, and then render the rest of the project with more or less instances.

  • Carlos
    over 4 years ago

    I rendered two animations (1 min each) and I saved a lot of money compared to a render farm. Barista would be perfect software if it could bake textures, as Blender is very slow for baking, I don't know render farms that can bake in Blender, anyway.
    From my experience, it's worth the money, as I saved much more money.

    • CG Cookie

      about 4 years ago

      Thanks Carlos, and definitely!

      There are always plans to make Barista better in so many ways, and an option to make Baking a possibility is one of them. 😉

      Happy rendering!

  • Mutiny FX
    over 4 years ago

    I work at a small VFX studio and we recently started trying to use Barista as our online render manager. It has a lot of promise, but some missing features and technical problems keep it from being great.

    We have, in some cases, needed to render a custom frame list(eg 0, 21, 48, ex.t). There is a slightly hidden function for it, but It may be hidden because its a little dubious. It always skips frame 0 (No matter what I do), sometimes it skips the first frame in the list, sometimes it skips random frames throughout the list. It requires a fair amount of babysitting just to make sure its rendering the frames that were submitted.

    As far as I have been able to find, barista only allows you to submit one shot at a given time. This is a problem if you are missing a frame (lets say frame zero) and want to queue it to render in parallel with the other frames. You can't, instead you need to wait for everything else to finish and then start one instance and queue your missing frame by itself and wait till it finishes. It would also be super useful to be able to create a queue of multiple shots to render during the night or weekend.

    Its also sometimes just disconnects from AWS and can't submit any more projects. However it can usually still start new instances, so I believe this is just a glitch.

    All in all its a very useful bit of software, if you a just dealing with shots where you can use the from-to frame submission its far more stable. It also automates a lot of the more technical side of dealing with AWS, making it just a few clicks and then you have loads of on demand processing power. For single shot renders this is a fantastic tool.

    Disclaimer, this was all written about v1.4.2. I've downloaded v1.4.3 but haven't had a chance to use it yet.

    • CG Cookie

      over 4 years ago

      Thanks for the feedback!

      The whole reason this app exists, is to provide a comfortable rendering experience for everyone, but no one shares the same render process. If you are experiencing an issue, we would love to hear about it! If we don't know about your use case, we can't make it better for your specific needs.

      Prior to 1.3.0 you couldn't even use custom frames, but then a few customers mentioned how they would love the ability to render a custom list of frames, so we added the custom frame feature a week later. All you have to do is let us know your use-case, so we are aware of it, and we will figure out some way to make it work in your favor.

      While it doesn't natively support multiple renders at this time, other customers have been launching multiple clients to accomplish the same thing, and it works very well. While this method is not officially supported, we do have plans to add official support in the very near future.

      In regards to your other issues, I would really like to know more about the problem with it "disconnecting from AWS", as this is the first time hearing about such an issue. If you would please contact us through support, telling us more about this issue, we would really appreciate it!

      We always welcome feedback like this, so if there is anything else you haven't mentioned, send it our way. 🍪

      -Dennis
      Barista Development

  • Morgan Garcia-Lamarca
    over 4 years ago

    Definitely the way to go for cranking out renders. Depending on the render times/volumes it can be cheaper per image to more less expensive machines vs. a few more expensive machines. I had an issue accessing the GPU instance list, and got a response and fix the next business day. Great service, and great potential for saving time on renders.
    Great!

Purchase this item to leave a review!

Choose a product version:

  • $69.95 Complete

    Access to both stable and development versions

Sales 300+
Customer Ratings 21
Average Rating
Dev Fund Contributor
Published over 5 years ago
Blender Version 2.6x, 2.7x, 2.77, 2.78, 2.79, 2.8, 2.81, 2.82, 2.83, 2.9, 2.91, 2.92, 2.93, 3.0, 3.1, 3.2, 3.3
License Mit
Have questions before purchasing?

Contact the Creator with your questions right now.

Login to Message