Powell Dobson Visualisation were asked to produce a 90 Second animation for the exciting ‘Afan Valley Adventure Resort’ in South Wales, UK. Head of Visualisation Will Morris tells us how they did it.
The client gave us a very tight three-week deadline to design and produce an animation of their four different themed lodges types to show off to potential investors. With a team of three I knew it would take us almost three weeks to complete all the modelling. This timeframe left no time to render the animation in-house; it would take roughly 2-3 weeks on our internal render farm. With 2250 frames at roughly 50mins per frame (dual Xeon) at 1080p, the only way to render the animation was through a cloud render solution or online render farm. This was a first time experience for me, and something I have always been nervous of. This breakdown will talk through how we created one of the scenes and what we had to do to control render times, whilst ensuring it rendered glitch free.
Firstly, we started by modelling the building (all done in 3ds Max with basic modelling techniques). For this particular project we didn’t have any CAD elevations: only a CAD plan so we imported the plan into Max and pulled the walls up and designed the lodges with our in-house architects as we went along. We also modelled the surrounding ground/terrain; starting off with a flat plane and pushing and pulling to give an uneven rural looking terrain.
The next stage was setting up the animated cameras. This lodge animation consisted of 5 short 6 second (150 Frame) sequences. With limited time on the project, we didn’t want to spend time modelling something that wouldn’t appear in the animation. So, by setting up the cameras early and agreeing them with the client we only had to model what appeared in the particular view.
As each camera sequence is only 6 seconds (150 Frames), we only wanted to move the camera a short distance which helped to ensure that the camera movement was smooth and slow. In the ‘Time Configuration’ settings we set the frame rate to ‘PAL’ and set the animation to 150 frames which changes the time bar along the bottom. Then to animate the camera, we clicked ‘Auto-key’, on the time bar and slid the marker to frame 150 and then moved the camera. This recorded the camera movement.
We then added 3D furniture and design to the interior spaces. Due to time constraints, we could not model from scratch, so we downloaded the majority of furniture from 3dsky.org
We used a Vraydome light with a HDRI to light the scene and controlled the exposure values with the physical camera.
We then created the surrounding landscaping and vegetation. As this was ‘Alpine lodge’, the client wanted to give it a wintry feel. We used some wintry looking trees from Evermotion and Grass from R&D Group. We looked to Forest Pack Pro to easily Scatter all the trees and grass. Below are all the models we used to create the landscape. To create the landscape we created a number of different scatter groups:
One of the most important things we had to do was optimise the scene to render the animation; the challenge was getting a good balance of Quality vs Render times. We had 2250 frames to render and the longer the render times the more expensive it would be to render it externally. Below are the most important things we always consider when rendering any animation:
The longer the frame times the more expensive it is going to be to render it online. Reducing frame times by 5 minutes per frame can have a huge cost saving. It is all about getting a good balance between render quality and frame times. From the minute we finished modelling we had 16 individual sequences to render in 2/3 days with 2250 frames at an average 50 minutes per frame. That equates to 1,875 render hours on a single machine.
The render settings for the animation were pretty much left as the V-Ray defaults. The are two methods of rendering, both of which I have used successfully. You can either use Irradiance Map and Light Cache to calculate the GI: you will need to pre-calculate both before rendering the frames. This method is more fiddly and time consuming to set up and you are more prone to make a mistake (e.g forget to tick an important setting) but I find the render times are quicker if you are rendering an animation where only the camera moves.
The 2nd method and the one we adopted for this animation is to use Brute force and Light cache, this method requires almost no set up time, you literally just hit the render button and you get a more accurate rendered result without any worry of errors. The only problem with this method is the render times are higher. As I only had 2 days to render the animation and I did not have time to recalculate the GI for all of the individual sequences, we chose the brute force method. I probably could have brought the cost down of the animation if we had used irradiance map; for speed of set up and peace of mind brute force was the way to go.
When using Brute force, the most important setting which controls the quality vs render time is ‘Noise Threshold’ under the ‘image sampler’ roll out. The lower the noise threshold the better the quality image but the higher the render times. After a number of test renders trying out different values I ended up using the noise value of 0.02, this gave us a good balance of quality vs render time.
Now the scene is ready to render, we had to do one last set of checks to ensure that there are no problems when submitting the job to an online render farm.
After we had fully optimised my scene and done all of our checks to ensure the animation will render problem free, it was then time to render it. As mentioned earlier I had never used a render farm before and we were extremely nervous using one; there is always the worry that I could submit something incorrectly and it could all come back wrong without knowing how much support would be available to me.
I came across YellowDog on the approved list of online render farms on the Chaos Group website. The support from the team was incredible; they were available 24/7 and this immediately calmed my nerves. Their simple 3ds Max plug-in was brilliant for submitting jobs to render. When we were ready to render we would click ‘Export’ and the scene would immediately upload to their server and begin rendering. For every sequence, we would submit a number of test frames so that I could check that everything was rendering as it should have been.
YellowDog’s brilliant online interface allowed me to log in and monitor my frames as they rendered which gave me peace of mind as I could check the animation quality in almost real time.
YellowDog rendered the 2250 frames at an average of 50 minutes per frame in the time that I required which gave us a full day to composite all the frames in After Effects and send to our client. The animation has been very well received.
You are seeing this because you are using a browser that is not supported. The YellowDog website is built using modern technology and standards. We recommend upgrading your browser with one of the following to properly view our website:Windows
Please note that this is not an exhaustive list of browsers. We also do not intend to recommend a particular manufacturer's browser over another's; only to suggest upgrading to a browser version that is compliant with current standards to give you the best and most secure browsing experience.