 Good afternoon Blender Conference and thank you for welcoming us. My name is Pierre Andreani, I'm the manager at Tata Motors in the UK. I spent 12 years at General Motors in Detroit before becoming the manager here. In December 2014, Matilde graduated from her school and this is when I gave her her first interview in November of 2014. And this is when the first time I've seen a Blender in action. So the rendering of the helicopter was really the first time I ever seen a Blender and I was impressed at the time. In January, before her second interview, I had the seats that had to be modeled in-house and I was debating if I should use polygonal modeling. Knowing that Matilde used Blender, I was wondering if we could model those seats in Blender. So she did a test for me and after three to four hours, she finished the seats. And that really was for me the first tell-tale sign that we should use Blender. So after the first test of the seats, we went all the way to full adoption of Blender in about four months and Matilde will detail all these steps in her presentation. About Matilde, as I said, she has helped us with Blender. She has modeled the car, she also modeled the organic figures such as self-portrait. She's quite the expert in modeling, figure modeling, lighting, texturing and she will detail the outline of Project X that we did in-house. Once again, thank you for welcoming us. We're really happy to be there and I let Matilde walk you through our presentation. Thank you. The talk of my manager basically. So I will start by introducing my company a little bit of history and a bit of the future. Then I will speak about the process. How we integrated Blender in our process. Then how we see the future of our process with Blender and we'll finish by some question and reply and things like this. So first, Tata Motors European Technical Centre, so TMTC to be short. As now 10 years old, we are based in UK in Coventry and we are fully on company by Tata Motors. We are doing some design research and development. So we are part of three design studios in Tata Motors. So the two other ones are the Tata Motors Design Pune in India and Trelix in Italy. In UK, we are mainly doing some city cars and some advanced creations, concept cars. In 2017, we will be relocated in the National Automotive Innovation Centre. So that will be a new building and we will be in partnership with Jaguar Land Rover. We are a manufacturing group and the UK government. So now I will speak a little bit more about our process. So basically that's our process. The thing we wanted to change is everything about the sketch modeling. So basically to be short, we are starting with a sketch from the designers. Then we are doing some digital modeling of this sketch to go with six different models. Then we choose three, then we choose one and then we produce this one. So the main part we needed to change was the sketch modeling part because it's the longest part in our process. So that's an example of a sketch. The next one is the concept car from 2014. And based on this sketch, we are doing some first modifications. So before Blender, we used the nerves. So we are using alias for doing nerves, alias from Autodesk. The best point for the nerves part, we can produce some very high quality surfacing. And we can have some very precise model but it's very hard to do some modifications. So that means that it's not a very good way to do some sketch models. That's why we moved until polygonal modeling and until Blender to be able to do a lot of modifications, some iterations. And then the data we export from Blender to the engineers are very light compared to everything we can get from alias or from the scan data that we have from the clay modelers. So that's physical modelers. But we don't have enough precision and enough quality surfacing to do all our process in polygonal modeling. That's why at one stage we are moving back into alias to have some very high quality surfacing. So we move until the polygonal modeling for a gain of speed and to do some very critical and very fast iterations that wasn't achievable with the nerves. And we move until Blender because we didn't have enough budget to be honest. It has no cost for us and no cost for doing some try if you will buy a license that costs too much and you do a try. And then you just don't want to do that anymore. That's a very expensive try. And I know that you said that Blender is not intuitive. I will say that Blender is intuitive. And when I arrived in my company I was the only one who can use Blender. And now we have all the team can use Blender. And the team get it very speed on it. That was very good. The main point for us for using Blender is when the head of the design exterior just come and say, we will do the change in Blender because it's faster to do that in Blender than I can show you on Photoshop. That's a very good point for us. So basically the sketch modeling is we have a sketch or a scan model that's coming from the designers. And we will have to rebuild it then doing some modifications on it or doing some iterations. We are giving this data to the engineers. They are doing some analysis and giving us some criteria. And then after we are sending the model until the milling or to VRED to do some virtual presentations, real-time presentations. And we are redoing these steps all the time along the designer is not happy with the models. And when the designer is happy with the model, we are switching until Alias to have some high-quality surfacing. So when we have some high-quality, when we have, when we finish with the sketch modeling, we are switching until the go for three. So as I said before, we have six different designs and we are choosing three. We are doing some more modifications. So we are reusing Blender for that doing some modifications and then switching to Alias to get some high-quality surfacing. Then we choose one design doing a little bit of modifications. And when we are happy, we are going to class A. So that's very high-quality surfacing. And that is only do with Alias because it's very accurate. At the same time, we are doing some real-time presentation, some virtual renderings and animations. So for now, these steps are, we are doing these steps with VRED from Autodesk. So basically, the creation of a car digitally is split in three steps. The advanced surfacing, advanced production and the production surfacing. At the time, this is split in two teams. And I'm a part of the team who are working on the advanced and the early stage of the production surfacing. So now I won't show you, but I will explain how we integrate Blender in our process. That was for a project that I can't show you. If you want to be alive. So basically, it took us four months to integrate Blender in our process until the time I arrived in the company to June by doing a first step of rebuilding a scan, testing with the engineers if they were happy with the data and doing some iterations. And then in June, we get the validation of the exterior designer and the interior designer to integrate Blender in our process. And proof of concept that Blender can work in big company and in design studio. So basically, we did a base model. Then we did nine iterations of this car. So with some different wheelbays and different lengths, different height. And to speak about times because we chose Blender because it was the polygonal modeling because it was a gain of time. For doing the ten variants of cars, it took us five days with two or three models because that could take us 10 or 15 days with two or three models with areas. So one week against three weeks of work. That's a lot. So now I will show you an example that I did just for the conferences to be a little bit more precise in what I speak about. The next one, that's the show car of 2014 that I will modify for an example how to transform a show car until a production car. So that's just an example I did it without a designer just by myself. So why not increasing the roof height, increasing the front overhang or some stuff like that. So basically I first start by rebuilding the scan mesh with a very low poly model that we really need a low poly model because we need to modify it after. So if there are too many polygons it's too complicated to modify. And as you can see here by snapping the vertex on the scan, when we smooth it it's just too small because we don't have enough vertex. So we have to fit the new model to the scan. So that took us a very long time to do. Normally we don't really have to spend all this time to fit our new model on the scan. So that's an important part we have to consider. And then when we have this new model we can try some modifications on it. So I will show you directly on the software. So the gray one is the scan model that we get from the clay modelers. For this part like it's a car that already exists. That's the LS model that I exported on Blender. And the blue one is the first model I did. So that's the one that doesn't fit to the scan. Then we rebuild, we fit the model to the scan and we have this car. So that's not very accurate as a model. We don't really have all the details or all the informations. But once again it's just a model for the shape. And we don't really need so many details on it in the early stage of the production. Because we just want to judge the design. So that's the production car. That's the show car fitted in the scan with a little bit of modifications. So as I said increase the roof height, increase the front overhang and things like this. Get the rear screen a little bit more vertical and things like this. And have an example of what could be a production car. But then we can repeat this step as long as we want and do some more iterations. And doing why not a forward drive car. So that's basically what we did for the project that I can't show you. We did one base model and then after created nine iterations of this model. And things we are doing with Blender 2 ways. Sometimes we have fun a little bit. And we are doing some crazy car that are not really cars. So what's next for us at the time motors with Blender? So there are still some steps that takes a long time or something we are not really happy with. Because we are used to use alias and we are used to high quality surfacing. So we will be very happy if we can avoid the steps of fitting the model to the scan. And we will be very happy if we can take the quality of surfacing, the reflections and things like this. There is already some madcap shading already exists. But not as accurate as we can have in alias. A proportionality thing that doesn't take an area but takes the points we want. As you can see here, I just took this point and asked him to modify the three points after. But not the area, the three points around the point I get. And we are still some problem of import and export files with the engineers and with the other software. So that's for now the things we would like to improve. And for later, we are thinking about integrating Blender for the real time and for the rendering and animations. Why not just thinking about the 3D headset, things like this. And if you have any other idea, just express yourself, tell me. And we will be very happy to do some tests and things like this. And so basically until the end of the year, we would like to try an animation of 30 seconds. Then we now using Blender just in UK. So we will start to roll out into India and into Italy. Then we will have all the time what I was using the same process. So Blender in the early stage and then alias. We want to come back next year to do a statue report. And if we are doing some stuff about, we will do some stuff about rendering and animation. Now we created a new process for that in the company because we don't really have any process now. And then in 2017, we will be relocated in the National Automotive in the NIAEC. I don't remember the name. Then we will be in partnership with GLR. So how can we work with GLR at Jaguar Land Rover? How can we get all the other companies involved in the Blender development? Be crazy. So thank you for your intention. And if you have any questions or if you have any advices or something like this, just feel free to express yourself. Thank you. I think this isn't so much time for questions. So maybe a very short one on that. Just a quick question as far as the things that you struggle with in Blender. Do you struggle more with the modeling tools or the surfacing quality? Do you find the modeling tools to be sufficient or are there particular tools from alias that would be really beneficial? I guess it's not really about the modeling tools. We don't know every modeling tools. I know that we bought Retropoflow, but we are not using it. I can explain why. Because when we want to use it, it's creating too much faces. Then we need to have a very, very low poly. So it's as fast as using Retropoflow that's creating the surfaces. But maybe in the next stage when we would like to have some more accurate models with Blender, we will definitely use it because I tested it and it's working very well. So the thing we are really struggling with is to get the modeling fitting the scan. That's the thing we are really struggling with. It's very important, I think you can understand that the designers are working very hard with the clay models who have the perfect body shape. We are doing that and when we smooth it, everything is smaller. It's like frustrating for them and we will have just taking time to get exactly the perfect shape to get something for them. And even not for the designers and even for the engineers because we are giving this data to the engineers. If the roof is something like, I don't know, even one millimeter less than the car, it can change a lot of things. So that's the thing we are really struggling with in modeling. Thank you. Thank you Matilda.