Datasets:

samfred2
/

ATO_TG

+{
+  "file": "3 - Selection of the Best Settings.wav",
+  "transcription": {
+    "audio_file": "3%20-%20Selection%20of%20the%20Best%20Settings.wav",
+    "text": "In this lesson, I will show you how the quantity affects the final quality of visualization. Samples, the type of denoiser used, the resolution of the visualization, as well as the light path settings. In the first example, we will start with the number of samples. I rendered several examples with different values. In each of these examples, the resolution was set. It was 1920 pixels on the longer side, which is full HD. We will start with 32 samples. You can pay attention to this wall, as the difference will be most visible here. And to the metal sheet, located on the kitchen wall. At this moment, we see 32 samples. Now I will show you what 100 samples look like. We can see that there is some difference. This quality is better. Now you will see 300 samples. Then 600. And 1000. As you can see above, a certain threshold. The difference is already so insignificant that it can be considered negligible. I would completely reject the idea of 300 or 400. These two values, or values between 300 and 400, are generally optimal. When it comes to the appropriate quality and render time. As for the time, 32 samples rendered in 6 seconds, 100 samples rendered in 17 seconds, 300 samples rendered in 50 seconds, 600 samples in 100 seconds, because the difference between 6 and 161 is significant. If we translate this into real values, 6 seconds is 6 seconds, and 161 seconds is just over 2 minutes. Then we have to ask ourselves whether the real value of the real value of the real value of the real value of the real value of the real value of the real If we were comparing 6 seconds with 10 minutes, then okay, the value would be quite large. The difference would be significant. It would be a chasm. But 2.5 minutes is a completely acceptable time. The next comparison will concern the denoiser. Regarding the denoiser, I rendered examples at 100 and 300 samples. In the first case, I used the denoiser that we set from the cyclist settings. In the second case, I used the denoiser that we can set in the compositing position. The denoiser set in this position looks like this. It is a specifically prepared node called the denoise node, to which the appropriately chosen rendered layers are then connected. If you run the program for the first time, you won't have such visible rendered layers. To activate them, you need to go to the View Layer tab and then enable everything in the light range and turn on denoising data. Thanks to this, the render layer window will be expanded to this form. After importing the denoiser node, you will connect all the elements together and then connect the denoiser node to the composite. For comparison, if we use the denoiser set from the sightless position, which is out, this one here. the using the denoiser. In this form obviously results in better outcomes, but it takes more time. Additionally, if we want to conduct post-production in Blender, and in this course we will do so, using this denoiser will be completely impractical and will take a lot of time. For this reason in the course I will use the denoiser from the cycles position. I will show you how it looks, for example. First we see a render done with 100 samples using the denoiser set from the cycles position. We will call it denoiser number, POPPA. Now I will show you an example using the denoiser from the composite position, which is denoiser number 2. Focus on this specific part of the render here. See that the difference is quite significant. Denoiser number 2 definitely extracts more details when it comes to texture. This texture is more distinct. It doesn't lose as much detail as with denoiser number 1. We can observe the same thing, for example, on the material of the sofa. In the foreground, denoiser number 2 gives quite nice results. Well, denoiser number 1 loses some detail. This sofa loses some detail when using this denoiser. Above we will see an example with 300 samples. It can be in the same place. And 300 samples when it comes to denoiser number. The effect is exactly the same. All of this was also rendered here in 1920p on the longer side. As for the time in the case of with two samples. The render time on denoiser number one was approximately 17 seconds. On denoiser number two it was 38 seconds respectively. Twice as long. As for 300 samples, these differences were 50 seconds to 69 seconds. Again a good place for comparison. Here we see 4K, rendered with 300 samples using Denoiser number one. Now I will show you denoiser number 2. We see that the texture is certainly very much sharper once again, but this difference is not as very clearly and noticeably drastic as with lower resolutions. A similar effect occurs on the sofa. The difference is of course noticeable, but it is not that significant. The difference is not as large as with lower resolutions. One can conclude that the higher the resolution, the more the differences between denoises blur. If we assume that the resolution we will be working with will be even higher, from 4k onwards, we can comfortably start using denoiser number one with quite acceptable results. I bet that if I hadn't shown you the two examples side by side, you wouldn't even notice the difference. You wouldn't notice that something could be more vivid or sharper. The quality is perfectly fine. As for the rendering time, denoiser number one took us about 3.5 minutes and denoiser number two took us just under 5 minutes. So the difference is roughly 1.5 minutes. In absolute terms this might seem like a lot, but in real terms, meaning from 3.5 minutes to 5 minutes, it's still a very short rendering time. At the end we'll compare the light path values. Here I have also prepared a few options. Let's start with the fast-gear option. Here the rendering time was one of the shortest... In the case of FESG it was 41 seconds. In the case of default GI settings, we see that the quality is definitely different. It's definitely better. The entire render becomes somewhat more, somewhat brighter. And we see that our shadows behave much, much better. The difference between fast GI and default GI was approximately 7 seconds, because default GI took approximately 48 seconds. In full GI, it operates with slightly brighter shadows and slightly better quality in detail. Here, the time difference between default and full was 2 seconds, because in full GI it rendered in 50 seconds. I remind you that the values in full GI have a value of 32 set in all their parameters. I rendered everything as a test with much higher values, namely 128. And here, there is absolutely no difference in quality, nor was there any difference in render time. We can therefore confidently assume that rendering using full GI is optimal rendering, both in terms of quality and time. Finally, I rendered everything for comparison with abstractly small values, namely values of two. And it looks like this. It doesn't resemble at all the visualization and quality we want to achieve. We see that something completely strange is happening in the reflections in the windows. They behave like glass. This is completely not what we want, and it is completely unacceptable. Although the time difference was quite significant when we talk about absolute values because it rendered in 39 seconds. So 39 compared to 50 seconds is an 11 second saving. But again, let's translate this into real values. 11 seconds next to nothing, a completely negligible value. Considering everything I have mentioned, we can conclude that the best option for light path lights is to render using full GI settings. As for final renders, it is best to render them in 4K. I think that using Denoiser number one, if we want to work on post-production in Blender, and if we want to work on post-production in external programs like Affinity Photo or Photoshop. It's worth considering Denoiser number two. It will be included as an add-on to the files for this course. In test visualizations we can easily use Denoiser number one because these are test visualizations so whether the texture is a bit more blurred or a bit less is a secondary matter in test visualizations. The point is to see the entire shape of the scene, to see how it is lit, and to spot any potential errors that need to be corrected. So here we can comfortably accept values, I think even appropriately. We can set the resolution to full HD, which is 1920 pixels on the longer side, and use denoiser number one. That's everything I wanted to convey to you regarding the significant impact and details of various different setting values on the overall quality and rendering times.",
+    "language": "en",
+    "duration": 707.77,
+    "timestamp": "2025-12-10T17:35:54.058910"
+  },
+  "timestamp": "2025-12-10T17:35:54.064242",
+  "processing_time_seconds": 244.05695056915283
+}