The Making of Divinity: Original Sin 2

I guess we get started good morning everybody thank you for coming my name is spending I'm the founder and director of large studios and so today I'll tell you a little bit about the making of divinity original sin 2 and the bruises we obtained while we were making it to start I'll give you a very very quick history of Larian just to give you some context for the decisions we made and the environment in which things were taking place we'd be making RPGs since 97 and our history can be split in two parts there's the part before 2010 and there's a part after 2010 and the part before 2010 was the part where we dependent on external funding to fund our projects so that those were publishers but it was also work for hire and post 2010 we took our fate in our own hands and we started making we started self-publishing our games and so Lehren has always been a company that tried to reach for the sky and often crashed and that's where all

the bruises come from but we never gave up we kept on trying and that has led to I have to say a very successful last five years but we feel we earned it because the first 15 years were very complicated so in 2014 we launched a game called divinity original sin one was very successful was a Kickstarter game I were asked 1 million dollars on Kickstarter and game sold really well had a high Metacritic was game sports game of the air it sold I think 2.5 million units by now so we're very happy about that and it essentially gave us the rocket fuel that we needed to finally reach for the sky and not try to crash on the ground but before we could do that we had to solve a few problems so after original sin 1 we came with the brilliant idea of making original sin - it was a very hard decision to make and we said this one is going to be the big one this is going to be the most

ambitious game that we've ever made and we are going to show the world that we always had it in ourselves to make brilliant RPGs to do that we needed a much bigger team and one of the problems that we have been having was that the cost of living and of wages in the country where we started which was Belgium was very high and when we look to for instance to the east we saw that some of our colleagues had wages that were almost half of what we were paying so that basically meant that they could have to developers for us one and another problem that we've been struggling with was that of crunch as we were reaching for the sky we often have to fix problems that appeared out of thin air and we needed to fix them and we were small team so that made that meant that we had to crunch lot so we try to get rid of that and another problem that we had to solve was we needed access to senior talents because in Belgium there's not much of a mature games industry at least back then there wasn't and we we needed people that actually had done productions had shipped many games and we couldn't find them or if we found them it was very

expensive to bring them over so the answer was obvious we started looking outwards to other countries and we'd been working with a small studio and st. Petersburg in Russia who had helped us with the Mac port of divinity or Johnson one and we approached them you ask them would you guys mind growing a little bit because we could use some help and they said yes and so that's how learn st. Petersburg was born and then we started thinking about how can we solve this crunch problem and the idea popped up that if you were capable of organizing your processes one way or another that you could basically offload the work that she had finished at five o'clock in the evening and send it to the other another time zone and they could start because for them it was the morning working on that maybe you could have a longer production line per day without having people to work continuously and that's how Lehren Quebec was born and then we had one last problem which I didn't mention that we needed Sophos we need a lot of writers English is not my native language as you can hear and it

was hard to find writers for us so we started looking at the UK which was the closest country where people were speaking English and we ended up in Dublin where we started a small office with nine writers and where we also decided to put our publishing team because we were self-publishing and like this we ended up with four Studios out of a small studio of 40 people in within the space of a year so all of this happens between September 2014 and February 2015 and so while the Quebec and Dublin teams were being set up the team in Ghent and the team in st. Petersburg started working on the vanity original sin one the enhanced edition for ps4 and Xbox one and so the idea was that each of these teams that we that we set up were basically going to be clones or mirrors of each other so that you could literally pass work on as the Sun rose and that was not the easiest thing

in the world to set up and it took a lot of time but it started working and we'll see some examples of that so by May 2015 we had teams that were operational in Quebec and Dublin the guys in Ghent as in Petersburg were working on the Enhanced Edition and so it was time to start working on original sin - and so it was the Quebec team who had actually never worked on an original sin game and the Dublin team which was full of new writers that were going to lead the charge and they started working on a small demo which was produced in three to four months which was going to showcase the pillars of divinity original sin - and which we were going to use to show to press - the community what we wanted to make and which we wanted to also take to Kickstarter and so the idea was that when the teams that were working on original sin 1 for ps4 and X 1 were finished they would then roll into the production that had been set up with the origin sin to demo so the core pillars of original sin - word that we wanted to have a better story that we got a lot of flack for the story

in original sin 1 so we were frustrated by that and so we wanted to fix that we also wanted to improve the combat we got a lot of flack over the combat especially for more hardcore community after the beginning of the game so as we went into Act two in act 3 we also found that our modding plans that we had for original sin 1 hadn't been that successful and dominantly because we hadn't really supported well so we wanted to fix that we also wanted to have a higher degree of polish because we thought that the gameplay that we were presenting could be consumed by a lot more players than actually had play original sin 1 and multiplayer was one of the big innovation original sin one in the art in the narrative RPG space so we wanted to bring more multiplayer options and so all of these things translated in the following big axis of development we needed a more gripping story one that was going to captivate you all the way until the end we needed a new style of writing so that you could people would see okay well they mean it when they're talking about a better story we

introduced what became known as the origin stories which was essentially the ability to roleplay characters who had a set background a set story a different set of quests within the world and we introduced a tax system which was essentially if you played Dungeons and Dragons and you know what the background is it was essentially a background that you got in the game and then whenever you talk to somebody they would recognize that you had that background and they would react as if you were that person and this were these were very important features that we were introducing to those two another thing we wanted to do is we felt that an original sin 2 or actionpoint economy was ruined because we had way too much action points so we wanted to be much tighter and this came from a philosophy that we wanted each point anything that you every single level a point that you used to feel important so that you would care about it and we also wanted to change the armor system over which we got an enormous amount of flack we'll talk about that a bit later the armor

system which was very unclear in original sin 1 and which we wanted to make a lot clearer in original sin 2 we also wanted to tie the story to the narrative mechanics that were being produced in the game so we felt the game was so large that you needed to have continued a continuous stream of new mechanics coming towards you to motivate you to continue playing and we wanted those mechanics to be present at every single spot in the game world and indeed in the story and then to top it off we wanted to have PvP multiplayer at least the option of having PvP multiplayer while you were playing the game and this came from the idea that the the core story was going to be something that was similar to Highlander in which there could be only one although we didn't know at that particular time what the only one was going to be you will see that that was also a big problem and then finally we wanted to have a updated graphics obviously a more realistic visual style and as I said we wanted to take modding into account from the get-go because adding modding after the fact is very complicated but if you

take it into your development PI point when you start in general it should work so the Kickstarter demo was really successful we made our million dollars in 12 hours or in the first day sorry and we reached our goal within 12 hours and so we were excited by that because obviously it meant that the direction that we had shown to our community was appreciated and they wanted us to continue in that role on that road sorry and because we we we hadn't expected it was going to go that fast we really hadn't prepared our stretch goals properly but we said you know what we'll just take those core pillars and will build our stretch goals around that and so this is how these stretch goals came to be so if you look you see love and hate the whole of echos racial skills the undead the shape-shifting mask they were all part of this better story ambition and his ambition of tying narrative mechanics to the actual story then obviously strategist mode and the extra skills trees those fed into the combat ambitions and modding and game master

mode they were all there to supply the the multiplayer vision that we had so the future was bright original sin Enhanced Edition definitive edition a story isn't nice Christ enhanced edition released it was very successful on ps4 and x1 we had money the the coffers were loaded we're a team we had four Studios we had a direction in the form of a demo which was called to prospect demo which was the very first version of original sin 2 so it looked as if nothing could go wrong all right so let's start with the story which is arguably probably the the thing that went wrong the most throughout the entire development and had an impact on and tremendous amount of things so these are actually the very first origin stories that we had these were present in the prospect point demo so the one that we use for Kickstarter and that we demonstrated to multiple journalists and which we used to show if you were that person suddenly all of the dialogues were going to change in the game because you became that person for the game

world now original sin one had very little character development especially when it came to the companions and we isolated this as a very real problem so the brief was for each character in original sin - you needed to have what we called a fear loss betrayal or wound you needed to have a mask or a way of coping with that a mask for this fear loss betrayal or wounds or a compensation mechanism for this fear loss betrayal or wounds and that was the thing that was going to grow throughout the game and this would give a gift depth to the characters we wanted each character to have course and eccentricities they had to have interesting aesthetics they need to have a good back story one that you would understand in just one line they needed to have motivations to go into conflict with the other characters in the game because we had this PvP ambition and they needed to have a story that was properly formed along according to enact one two and three structure now it took us quite some time to develop these origins stories and there were many reasons we'll talk about them one of the

biggest problems as the heads up was the fact that we had to make this work in multiplayer and if you've played the original sin - you know that we have drop and drop out multiplayer so that means that and sorry I forgot one thing in the original sin - you could also we had this ambition that if you picked an origin story that was fine he became the origin story but if you didn't pick it you would encounter the origin stories that you didn't pick in the game world and you could recruit them which essentially meant that we had the first person and third person view on the same stories and we had to implement that and that led to interesting problems because we have drop and drop out multiplayer so that means that if you start playing with an avatar when you pick an origin story and you recruit a couple of companions and then you have a friend that joins in and he becomes one of those companions the story had to rotate it had become first person to third as from third person to first person which was pretty insane but cool so it took us until eight months before release until

we actually figured out how to do this we've tried and we tried and tried and we couldn't we couldn't understand why we're having such a hard time but it took a very long time because it turns out that to be able to make it work like it worked in originals into you need to do a lot of things and it was so many things that we kept on looking for for for simpler solutions until we actually accepted the fact that there is no simple solution we're gonna have to do it the hard way so one of the main problems we had with his origin stories was determining who were you going to be what was going to make you special and it was very easy to come up with a reason for an origin story to be paid special but it was very hard to make that work with a reason that was also going to apply to custom characters which did not have an origin story because when you created a character you could pick up either an origin story like the red prince or fame or whatever but you could also create a character and say you know what I'm gonna be a scholar and I'm going to be a jester so

and that's gonna be my role in the game and the game is going to have to make sure that I also feel special and I don't know why we have such a hard time but we had because it took us a long time I think one of the reasons was that we try to avoid any of the tropes so we wanted to be special and we had a team of nine writers so we figured with Nine Riders we're gonna figure out something awesome so I kept on giving people time trying to look for something in the end we ended up with the trope but it it worked well enough but it took it took very long to get this specialness in working in the game let's put it that way another problem that we had to learn is a company that's been built to work bottom up so that everybody owns part of the game and can can can come up with their IDs and the general tendency is that if somebody has a better idea than someone else will take the better ID and and other tendency is that we we try to make it better and better continuously and so that's why we listen to any ID and we started with his attitude when we

started making the core story personals into and so because we were distributed all over the globe we started relying heavily on Google Docs which has this awesome collaborative features and this awesome commenting threat feature and so in the beginning the story was open for everybody so we had the scripters we had the programmers we had the artists everybody was making comments about the story all these comments were conflicting with each other and we're very polite people and we answer to every single one of these comments so several months of answering on comments and not making any progress progress led to the decision that we kicked out everybody except the writers and then we started continuing to work on the story but we had nine writers and turns out that nine writers don't agree about anything either sir so they just kept on arguing and arguing and we kept on patching patching patching on top of the arguments just to please everybody and then when you read what the resultant was it was just it was it was it wasn't very good so finally the politeness went out of

the window we brought it to a very small group basically two persons and when we did that suddenly everything flowed very very rapidly obviously we'd built already a insane amount of stories and there's literally thousands and thousands and thousands of pages of core story that have been written so we had an enormous database to to take IDs from and it suddenly all clicked and as it all clicked we suddenly figured out oh my god this is how we have to do the origin stories and we suddenly had a solution unfortunately this was core story was finished ten months before release origin stories were finished eight to nine months before release system was invented eight to nine months a month before release so the fact that our narrative was taking so long the fact that we haven't figured out how to do the origin but moments obviously had an impact on the I forgot to click on the thing with with the timer I don't work a well I'll figure it out so I have

no idea how much time I've been taking so far so the narrative delays that we've incurred had an impact on the rest of the production of course and so we we started with the things that we were sure of and we said well we'll just fill in and we'll confirm this is ready you can start producing on this bit because that's not going to change any more and that's not going to change any more book of course it changed all the time and so the scripters who needed situations the scripts and the dialogue writers who needed dialects to write they got used to the fact like well let's implement this and then a month later they got an update and said well this changes and this changes and this changes so maybe you have to remove the situation you have to add to this situation it led to something that was very frustrating for them one of the problems are the bigger problems that we had was the way that we documented everything so you're seeing two documents here I think you're looking I can't see it on my screen but you're looking at you're looking at what was called a narrative

design doc we had the core story document which was about 100 pages and then we had a narrative design doc which talked about every synth situation that you found in the game and then you had a QA doc which basically contained the flow charts and all of the permutations that you need to do to be able to test the situation now we started with big ambitions all of this documentation and we put a lot of effort into it but of course every single time we change the story we had to propagate those changes and then obviously mistakes were being made so things were not making it into one of the documents and that led to the situation where QA starts testing something says there's nothing to do with what's actually in the game and giving it at the end of the game we had situations where we were running out of time documentation was not being updated so it essentially was one gigantic mess and took quite some time to clean it up so the interesting bit about this is that we were value very well aware of it we tried changing the process during production we felt we had months of discussion after production about how to

do it we tried something else it was a disaster and we are still iterating through the documentation process how to handle this this is not a very easy problem to solve we have something now that we think works but maybe I'll be here in two years and I'll tell you don't do this so we had invested quite a bit in increasing the size of our scripting team because on original sin one we saw that we were under capacitated and so we said you know what we have a bit of money so let's do poor sales and let's make have more scriptures than we're going to need obviously as we kept them changed in a story that margin became smaller and smaller and smaller and eventually we became actually under capacitated now the way that we had organized our scripting was we had later in production because initially we had much more maps but at a certain moment in production we had three main acts act 1 act 2 and actory and so we said okay well one studio is gonna work on x1 once it is gonna work on ACTU and once it is going to work on Act three and then

we're gonna review that and this way we're gonna make the game in parallel and we're gonna be very clever about this and we're gonna finish all of this in no time obviously it didn't work the reason why it didn't work was not because the Scriptures were doing a bad job but they weren't getting feedback because the one thing that we hadn't paralyzed was a review process so reviews were happening by a small group of people typically deletes and myself and we were we just didn't have enough time to do it all on top of that the different time zone suddenly started working against us because we had limited windows of time that we could review with the team in Quebec when we were in Ghent and we had limited windows of time that we could review when we were in Quebec with the guys that were in st. Petersburg and games so decentralization of the review process was very needed but it turned out that we we didn't have people inside of the company that that could do this and as a result we we suffered heavily we actually solved this problem at some point by bringing everybody together again on the same zone so act 1 so we

put all the scriptures on the same ax well not all of them but most of them on the same act and that made it possible to review the acts in a whole and that that accelerated things and it's probably the thing that saved the project and in the last year because otherwise we were not going to manage to give you an idea of what it means to have a broken feedback loop with not enough reviews a classic symptom was a situation that was made by one scriptor a situation that was made by another scriptor they were standing next to each other and they didn't realize what was happening in the other situation so somebody could say oh the Sun is nice and everything is shiny and next to it somebody was dying and so you felt it very often but this was one of the the consequences of not having that review sorted out properly so once we started realizing that we were in trouble and that we were not going to be able to make everything that we planned on we

had to make cuts originally original sin 2 had a map for the human lands one for the Dwarven lands once for the undead so Bay for each of the races that were present in the game and what you're looking at here if you play the original sin 2 is a map of the human lands which contains the city of arks driftwood and for joy all on one map which essentially became the entire game so that gives you an idea of the level of cuts that we had to do so each of these maps had a lot of the a lot of action going on and when we cut it we suddenly started cutting and what worse was supposed to be the real estate for the origin stories so that meant that we had to go and change our origin stories again so like this this this constant ping-ponging between adapting because to the fact that we were going to be late and having to make cuts had an impact on the story the fact that it had an impact on the story meant that the rest of the production was was going to be late and so we struggled quite a lot with that but eventually we managed to have it under control

another thing that added to distress if you want of development was that we had we had hired 9 writers but these people had not a lot of experience in interactive dialogues we only had a few people that had worked on the previous games and the rest came from TV from classic writing or from television and the idea was that the combination of all those writers was going to lead to something beautiful and it did but we underestimated how much training that we're going to need to actually implement all of the dialogue trees that would fulfill all the ambitions that we had and so this is an example of a dialogue tree the the tricky thing about this is that when you zoom in on one of those dialogue notes there's still a whole bunch of stuff happening inside of those are notes that needs to be filled in and so there were all the fights between wall fights discussions between our scripters and our writers the scripters typically set up a situation they put all the flags and conditions correct and then the writers were

supposed to expand on that and then typically they broke all of the flags and the conditions and then went back to the script er who complained that they broke everything and then it went back to the writer in the right said well this is not cool so I'll change it and then we came and said well we've changed the story can you please change the dialogue you know and so like this there was a bit of stress and then an interesting thing happens and at the beginning of 2017 this was nine months before release we said okay we're not going to be we're not gonna be ready because we originally planned on releasing in June 2017 and we were talking about what cuts we were gonna make to still manage to be ready but then another thing happened there was we we're a company that looks very heavily at the timeline of competitors of what they're going to be releasing one because we prefer to release in what's called the blue ocean we don't like to go in a red ocean where everybody's fighting with each other and there was a game shadow war which was going to be released and we were very afraid of that one because we analyzed

it as being a game that was going to talk to the same target audience and so we wanted to stay away from it and they were moving around with their release day so we're like and another thing was happening because there was other games that we fought that appealed to the same target audience and they were all announcing their release dates and they were all taking up all the spots everybody had eyes analyzed on the timeline as being blue oceans and said holy hell we need to say or release date if we don't said tell the people or release date they're gonna take every spot and then we're gonna end up going against someone so we announced that we're going to release in September 14th and that was all release date and lo and behold he saw all the people starting to move their release dates around us so that was good and shot of war had the release date two weeks later so that stressed us out so we move our release date because we're in trouble because we have too much work and what do we do we say hey we have more time

so why don't we voice record everything right which was not planned we wanted to but we said okay we just can't do it we don't have enough time because and the script is not ready and not nothing's coming out but why don't we voice record everything so we took a napkin and we started calculating and you said you know what I mean Act one is pretty much finished because it's out on early access and it's the people are enjoying it we pretty much have a good idea of what we're gonna be doing so we're not going to lose all that time the writers have figured out the tone of voice of the game so they're gonna manage to write so why don't we hire multiple studios to do the recording in and just finish the script as they're already recording and then feed them the lines let's make an automated system that's going to be able to handle that and should work right so a piece of cake and worst-case Plan B you always need a plan B we'll hire we all add a third or fourth studio how hard can it be so we

learned from those two that was important to her sorry does one that was important to have a automated pipeline for doing your voice recordings and so we we had a system we had some changes made to it and we said let's do a field test to see if we can actually handle this and the field test was it was successful you know the script was generated from a build server and that build server ran at night and then in the morning he had a voice script that we collected text from all kinds of files and you could send that to the voice studio and the voice studio at you IDs that they had to associate with the clips that they were recording and then they send they upload it to a server they send it back to us we imported it on the build server in the next day everything was inside of the game so seemed to be a very good system and we knew of course that if we were going to still continue to write on the game as they were recording there was going to be wastage but you know we said it's the the downsides are vastly outweighed by

the upsides of doing full voice on the entire game so here's an email that I'll let you read that we sent to [Music] sorry the email I have it's not email sorry excuse me so by the end of June we know I'm sorry I can't see it so this is my problem okay so this is the back of that this is the calculation that we did all right so we figured that we were gonna have at one finished in a certain time act - in a certain time actually in a certain time that we're going to be able to calculate everything and on June 28th we told to voice to do that the recording was going to have to be or something around around 600,000 words and on July 12 we had an emergency call with them to tell them why it's not 600,000 where it's gonna be a million words and to be precise it was 1 million

78 79 thousand five hundred sixty-two words which was a bit more than we bargained for and so you may wonder how the hell do you end up in a situation like that and that's the question we talked about a lot so it began with the word count now to explain everything that happened you need to understand how we design our situations so we have this concept which we call n plus 1 design and n plus 1 design basically means there's n permutation or n ways that you can solve a quest and then there's always a fallback situation the reason why we always have that fallback situation is because we have a very open game in which player agency is everything and that means that killing an NPCs for instance something that should be possible as a matter of fact you should always be able to kill any protagonist any antagonist in the game with very very very few exceptions and what that means is that even if you have a very nice story that you're trying to tell the player might come it might come in launch an Armageddon spell kill everything and you still have to be able

to tell your main story to the player and he expects to be able to continue throughout the game so n plus 1 design is a solution to that and our scriptures are expected to make every single situation bulletproof put armor on it make sure that it has all the requirements of n plus 1 design so scripters is obviously our humans and the as they are going through a default type line where the story is changing all that they start taking shortcuts they say you know what I'm gonna put this situation and I'll script it and then I'll just wait for them to be a hundred percent sure that the situation is fixed and then I'll put my inputs one design on it otherwise I'm just gonna lose way too much time and I know how this goes so by the time they started realizing oh they're actually serious they want to release the game and they actually know what the story is which is cool so we can actually finish our work let's put in the n plus one design permutations right like literally a nuclear explosion at the scripture because what what n

plus one design often means like let's copy paste copy paste they're gonna get a good day and but there's lines appearing everywhere ding ding ding ding ding ding ding and the scripture they would the way that we work is the scripture stops the line so he puts it between brackets and then the writer afterwards knows that if it's between brackets he needs to replace the line except we had no we had never made a dashboard that counted to the amount of stop lines so a lot of this was invisible for us we didn't see what was happening but it was happening and had producer that started noticing he first felt that his database was broken because he likes it's impossible I'm doing double accounts something like that and another reason why the number of words was exploding was and each of those notes if you remember that data design 3 that I showed you we had actually not one line of text we had multiple lines of text for each of the tags that the players could have so if you were a lizard he would have a line if you were a human you would have a line so and that descriptors had also postponed or the writers had postponed

so that was being added so that basically meant that often a line became nine lines all right and like this or working just as well went all over the place and yeah it's a it was problematic so that wasn't the only change the only reason why word counts kept on going up going up because maybe a dialogue had then finally been finalized and all the end +1 cases had been done but there was still somebody adding extra extra lines to the game because they were the reviewers sorry we were late with the reviews and because we were late with the reviews we still had feedback to give two descriptors of things that they had to change or feedback to gift the dinocras of things that they had to change and that too was adding to the word count or not necessarily to the actual word count of the full game but it meant that the lines needed to be rewritten so we had quite a lot of iteration going on and then yeah the inevitable happened that voice system that we had and that that localization

system that we feel tested it started breaking it wasn't made for such rapid iteration it couldn't handle so much change we kept on hacking it because we had this active that needed to be recorded so we changed this or this in the pipeline and eventually it went it's all like a pudding fell into one another and so we kept on patching and patching and patching it and it was quite the effort to maintain everything and we we certainly rerecorded all of lines more than we should have but we made progress and inside the room where all the riders thought you could see magic happening we could see where we were heading and we we could see that it was worth it but the rest of the team was getting very angry at us because why are you still changing it and why are how is it possible that even now you don't know all your lines so and there was another cause of extra words which is that we were still adding features we had until September right so we had this this was one of our stretch goals the undead and because our origin story system had been

so late and development we hadn't even been we hadn't been able to start on the undead stretch goal now the thing about that stretch goals that you need to understand as this was - a feature that involved multiple mechanics and it was to be the showcase of how systemic or narrative system was so to to explain that if you were a lizard and you talk to people they talk to you as if you were a lizard if you were a nun that they freak the hell out if they if they were afraid of him that and so we had to have different behavior the idea was that you were going to be able to handle this because you would have a mask of the shapeshifters the mask of the shapeshifter was cool because it changed your tax and it showed case that if you change tax the dialogues change it so they rotated before that was a very awesome feature because it showed exactly how deep the game was and we absolutely wanted to have it in there one thing that we hadn't fought off was that if you earn on that and there's jokes or there's lines that that talk about your face you would have to rewrite it also because you were an undead even in the

narrator note which was something that was very rare and so this here it was an example of the the past that was done over the entire game to be able to handle the undead and it turned out there was a lot of words describing the census and the the bodily structure of persons inside of the game so there was a lot of rewrite happening as a result of that also now we've been searching for the voice and the the tone of the story for more than a year and we finally figured out and we were all convinced at least in the end the narrative and writing team that we needed to apply this throughout the entire game because it was gonna make the game that much better so we fought with the producers heart to get all of this stuff and what you're looking at is what define my summer in 2017 and this is the file that was eventually generated somewhere and mid July it was

a file with 80,000 lines of text in it and more than a million words and the little ones that you see meant that something was ready to record and when it was the little one was put there it meant that the situation had been reviewed to dial I never reviewed it was okay were never gonna change it again in our lives which we didn't but that meant that between July and August as we were recording by now in multiple studios we just were going there one okay this is not okay I have changed this change let's change this and so it's fair to say that half original sin 2 was written in this couple of months as a result of all delays that we've been accumulating lucky for us we had a lot of writers and then we could handle it but it still it was an insane job it was a yawn or writing director and that to give you an eye an idea of the level of iteration that we applied on this game one week before release I said do you think I can still get this actress and we changed the

middle of the origin story of seville and indeed it was a lot better and then he asked me like how do i do this voice recording pipeline because he didn't want to go to the producer the script had changed so now imagine yeah so now I need I have this slide here because of the producers deserve a lot of respect for despite being grumpy and deservedly angry at us they did cover us at every single moment in time they made sure that everything works they figured out solutions for all of the problems that we encountered and it's you can for sure say that without them nothing was going to was going to happen it would have been one gigantic mess now imagine that you're working at learn studios and you're working in QA and you have a test plan and your test plan is changing the entire time because somebody's changing the story the entire time and even as you are like supposed to to test your finals they're still flagging this is ready oh this is not ready let's change

this so this is ready your job is almost impossible so our Q tech QA team was actually yeah they were amazing they had roadmaps that went through the game that tested all of the permutations and they did they did a pretty good job but it was arguably all it said needless to say not everything had been tested when we shipped because it was impossible Christians who kept on changing things continuously we did have some automation and that saved us we had well testers that basically went over the entire game automatically so there were the NPC's that we're doing combat automatically the all the levels were tested automatically the one thing we didn't have in something that we're investing in now is something that tested the story that tested the critical path so it needed to be played and the game was so large that if he started on Monday morning of one week then the Friday of the the week after he did one run-through of the game so he can imagine how much QA it took just to be

able to do a permutation and so as a result with have that much view of how the entire story played out because it just took too much time to go through it luckily for us it didn't matter that much that certain things didn't work because you couldn't necessarily notice when you play it again divinity had so much reactivity that if something didn't react you didn't know if it was supposed to react or not react there was only one thing where you could see it and that was in the journal and that's the one that really hurt us because if you looked at the journal a quest remained open forever and that was a clear indication that something wasn't working and that was it actually without that we would have had a 94 Metacritic instead of a 93 and we got a 7 out of 10 won 7 out of 10 from polygon rightfully so because she was on a yeah she had a really lots of problems in there and then her questing and that was actually a result of all of these changes so it's not something that I would recommend on the other hand if we wouldn't have done it then the game wouldn't have been as good and as well received as it was so

there was more going on than just a story but story was probably the biggest thing that that went wrong you're in original sin too that's why I spend so much time talking about it we also wanted to improve combat and so we wanted to improve multiple mechanics and so we decided the armor system should be better and we're very enthusiastic about it because we tested it we tested it at PAX and we saw that people instantly got how armor worked and it was very intuitive to them and so we were very enthusiastic when we launched it on early access except the community hated it they really hated it they wrote books about it and we were very stubborn we didn't want to change it because he says well they don't understand it so accessible everybody understands it blah and yeah lots and lots and lots of discussions there were an incredible amount of iterations inside of the company trying to save the armor system but essentially it remained a broken thing and it's something we shipped with and maybe we shouldn't have shipped with it in the way that we did the problem

with it was that if you combine a damage shield together with control immunity you you limits your design space tremendously and that's something that you don't really feel in the beginning of the game which is what we had in early access but you definitely feel it as you move towards act to an actor II another thing over which we traded a lot were the surfaces we in original sin one was the game in which you could there was already put fire on it you got an explosion there was water you put the interest on it you get an electrified surface you put a fire on water it was going to you were going to get a cloud a steam cloud people liked it a lot because you could involve the environment in combat so we figured that we should expand on this and so this is one of many many many flowcharts of systems that we tried to get the surfaces better and the thing was that and we had those things executed so or the effects team started putting the effects for it and we had like.we there was one level I couldn't find a screenshot back for this talk

where you had all four joy with all kinds of different surfaces and then the game started of course what does this surface do and that was the core problem of it people couldn't rely couldn't understand what was happening with the surface it wasn't one of the basic elements so eventually we dropped all of that and we stick we stuck to the surface system of originals in one and the only thing we added was were two adjectives blessed and cursed and even right to be honest even to this day I still don't know what cursed and let's do because they changed all the time it wasn't very clear so it was something we traded on a lot one of the biggest problems we had with early access was what was what was oh my god what was not on early access so we released act 1 on early access and the reasoning was we want to make sure that players when the game ships are going to have new content to play through otherwise there will not be any excitement if they already consumed all of the contents and we had people that kept on playing playing playing like they had already stacked up hundreds of hours just on the only very

early first act and so the problem was that we were getting a lot of data from the from for joy from the act one but we weren't getting any data from act two and three so to explain how we got into the problem that I'm about to show you one of our main goals with the systems design was that each stat point was going to matter and at the leveling up would be a very tangible thing so that you would have a continuous desire to improve your character and get better equipment and better skills and we had a world with the very few bottlenecks so the only way that we could guide you was through level scale no not level scaling level gaps so if you were level 2 when you fall a level 4 creature it had to be very hard but we still wanted you to have a chance to defeat that creature because if you manage to do that using the systemics of the game you were going to feel good about it so what we did is we had a very steep leveling curve so it was about 20% per level of extra damage

hit points and so forth and that meant that you you would feel it very very strongly when you when the level up but it also meant that if you were a level too high it was going to be very easy to kill something if it was at a lower level and that's the thing that that went wrong so this is data that we obtained from early access this is a heat map of where people were walking and this is a heat map of where people were dying and so that was that was really good but we didn't have this information for act 2 and act 3 and so what we did is we invited play testers to the company have volunteers to try things out and we we tracked the data from them the problem was that we couldn't find a lot of volunteers that were ready to spend two weeks in the studio because that was the time you needed to play through the entire game so the only data that we could capture was the data that we got from our internal QA and so our internal QA they know this game very very very well so to give themselves an extra challenge without telling us they started having

parties of three characters instead of four characters and they went through the game and then obviously they had a bit of a challenge and so what we never saw was something that became very apparent in the first week of release was that the game if you did everything accrued too much experience something that hadn't shown up in any of the tests that we would that we did before because huee was following their critical paths and we had some runs that tested everything but given the amount of time that it took to test all of it the data never really popped up game comes out or min max er start doing everything in the entire game by the time they're at the end of act 2 there are two levels higher than they're supposed to be by the time they're in act 3 they're way too powerful they're just breezing through it and say oh my god boring and that was completely against the design that we had where we wanted you to feel every single point that we were putting in any level so we panicked we literally panicked because we knew how important the the leveling curve was and we

intervened we released a patch four days later and which we hadn't tested and we just introduced spikes no so we just upped the difficulty level in the middle of act two at the end of Act one and at the end of Act two if I'm not mistaken so these are the spikes that you're seeing there I worked certainly worked again became harder and the reasoning was like okay these are the guys that that played through really fast all right there's a whole bunch of people that will play through a lot slower so we're going to test the leveling curve now and change it as much as possible so that we the people that are that are trailing if you want through the game are going to have a better experience and so over time we managed to to smooth it and round it and we solve the problems because there were some other problems going on also we got the everything sorted out but it was another reminder that if you don't visualize your data or you don't have enough data to make decisions on it becomes really really hard to to tweak your game

there's another change that came in very late AI 2.0 so we learned from those one that our combat was very slow and people take a lot of time figuring out what they're going to do and it's a lot of fun and to make sure that people wouldn't get bored as they played the game we try to make every single combats handcrafted and very different to the other combats that you were encountering in the game and so the prescribed approach was to make tactical puzzles out of each combat which you could solve in many many ways now you need to know that the approach that we have to the design of skills was to introduce concepts early on then teach players how to combo them and then later on give them some skills to top it off and what that meant for the AI was that it was very easy to foresee what players were going to have in the beginning of the game it became much harder in the middle of the game it became almost impossible at the end of the game because we had I think close to 300 skills and and it's also a a class this game so you can just mix and match any skill that you want so in the past

with original sin 1 we had a very simple primitive state machine where it was if-then rules that were made by combat designers that we're going to decide what the AI was going to do and that meant meant that if you figured out some kind of strategy that was successful against AI you could keep on doing it because probably the combat designer had not applied did the whole and the AI sinking throughout the entire game and so in October of 2016 so 3:9 a year before released we say you know we maybe should have a real AI now uncharacteristically so I was against it because I thought it was too risky that we weren't going to manage to get it done in time because there was going to be a lot of fiddling with weights and if the approach didn't work we had nothing and the other thing that I didn't like about is that it meant that we actually couldn't really finalize any of the combats in the game until the AI was going to be ready and they told me it was going to take two months but I didn't believe a word of it rightfully so so this was one of the

cases where or multi studio structures safe today we had programmer working in again we had one paired with him in Quebec and together they spent a 17-hour day and so they worked and worked and worked on the AI and by somewhere in january/february we started having something but they did work on own it's got like this until the very end of production but a 2.0 was fantastic it's really added to the gameplay that was present in the game there was a whatever strategy you came up with the AI could conclude it could could do something cool against it it was very clever also about abusing glitches in the system that we didn't even know that existed probably the most famous example was the bartender that drank all the alcohol to buff himself not dealing with a corner because it was in combats right so or we had an AI that shot a barrel of oil not sorry and I that shot the barrel of oil

to slow you while you were hasty to get rid of your hasted status and all these kinds of emergent behavior came from the AI so I think it was a very big element in success so I'm very happy that the programmers that argued for it won this particular argument so all of this problems and there were more but I am running out of time all of these problems obviously had an impact on our release and in our particular case we had multiple release days dates our very first release date was our Kickstarter because we we went to pax we went to journalists we released a demo that they could play and that's how we had all of our previews or second release day it was already access because then we went to the public and we said here's act one of the game that we're trying to make please play it and let us know what you think or third release date was internally called as Kirill release day because somebody had clicked on the button to release the game prematurely and this was two weeks before release

luckily luckily we had a phone number from somebody at valve that we could so directed that we could call to rectify it rapidly because other it would have been a disaster and then we had the actual release date which in our case was plagued by power outage in the entire city so we we had to do all kinds of stunts to get the game actually released and then we had the definitive edition release status when we came up with an improved version of the game which released last year so that was literally the fifth release date for those two now on release we paid the price for the word count and the thing is that on this one I think it's in the next one yeah we paid the price and that we were review bumped and we were review bumped by very angry people in Russia that said where the hell is my Russian localization that you promised and we said it was going to release on 21st of September because the translation companies have suffered some delays because somebody has been changing the script until yesterday

and and we haven't talked about translation but those guys were heroes also because it's theirs translation there's proofreading and when the game that's this big proofreading takes a long time also so and they were angry they were so angry and you could see it so what we did is we said you know what we'll just release the betta text right I mean like you'll see it's it's a really good translation it's just not hundred percent finished and that actually worked we went to twitch and in Russia we talked to Twitter's and to calm down the community and said like this is what they're doing they had a few problems and the sebuah and so forth it worked I would certainly not recommended because we were at that moment we had a 96% user rating on Steam and we were I think we went down to 70 or something like that so we were we were suffering very very very hard but it was our own fault of course because we changed so many things so if you hear all of this you would say well what a disaster right but there's a couple of things that you need to realize when you

think about the original sin to this is 120 our game that was made in two years time so that is I think quite an achievement of the team and it was a game that went really arrived because it's court very very highly it sold very well also it's still in the charts to this date so and it was a testimonial that that all of that is relation and all that of the the refusal of compromise on quality that it paid off even if it crossed a few production problems so my lesson was that not compromising on quality is a good thing but you need to figure out how the hell you're gonna make your production work with it because otherwise you're going to keep on running into problems and its problems that you don't want to have so what are the big takeaways that we've applied to our own processes on the next things that we're working on and which also not all work the first thing is documentation balance and documentation for a big RPG is really important you're gonna spend you're going to create thousands of pages of documentation for

QA for your scriptures even the briefs of the writers so you have to figure out a way of not having too much documentation because the last documentation you have the easier it is to maintain it on the other hand you have to be able to inform everybody about what's supposed to be going on so you need to have sufficient documentation this is a very very hard problem if anybody knows the perfect solution for it please come tell me white boxing white boxing refers to the how do you make a game that you can review without having to put all of the production values in there it turns out for a game like original sin - it's very complicated we've often observed that if a level design is not far enough a and decorated enough a script or it's not going to put efforts in making all of the permutations that he needs to do and that is yeah that is a very surprising thing we've fought against this but we never solved it so figuring out a good white boxing process and I think we have one now is very important Automation invest in automation as much as you can it is the saving grace when

you're iterating I mean because if you don't have to have a human going through everything and you have something that can automatically test your game you're gonna save so much time your processes need to be built for iteration you need to tell people up front this is what we're going to be doing we're gonna change until the end all right if you don't do it you will have a lot of frustrations you'll have a lot of fights and you can avoid a lot of these fights by just being very open about it yeah the too many cooks syndrome I think that's obvious you need to have some dictatorship involved in making a game but you still have to open up the line so that people can understand what you're saying feedback loops we talked about that invest heavily in it it's very important real world circumstances we saw a few examples of that and hard to organise but definitely worth it and but topic on its own never stop improving that's the thing that we that we really did well I think and it paid off for us and then you have to accept that iteration takes time and there's nothing to be done about it you're never going to get it right from the first

time if your production is completely production focused then you will lose opportunities to make your game better I think that when you make a game you should strive to make it as good as you can and sometimes you have to just take risks and they can they can blow up in your face I had one blow up in my face this morning actually more to be said about that probably in the next week's but yeah visualize visualize as much as you can make sure that you can see data that you can see disasters happening like the word come that we had and yeah so these were the top things that we we used to change your processes whether or not they work will see in an ex post mortem but that's going to be it for today so we have four minutes left for questions so I guess we jump to questions right away if there's any questions thank you this on high so I love this game I've

played it a lot I've noticed you didn't mention act 4 at all did that just like appear no so what we call act 1 2 3 is basically the three act structure the narrative one so multiple maps are actually involved in act two so Council of seven and what we call our C main which is where drift to attest those are all one act for us okay so the last act the one that you're referring to as acts 4 is actually act 3 ah so that's why thank you all right you're welcome good talk thank you so I have a question about the back door I noticed that in the gaming can have one character talking with the enemy while the other is putting something very heavy into the back Packer of that enemy and so later when the the fight start that guy could not move at all so it kind of I don't know why you have this back door is that deliberately designed because I don't believe he's simply a bug so could you explain why you have

that I mean there are other backdoors like if you shoot in in distance at a distance you would not you're like your movement would not be locked but you can especially using the the Sun or forest that character you can actually all summon earth with it on purpose we did it because we know that players like to glitch the game so we think is of a way of promoting player player agency if they figure it out they're very proud of it it's sitter it's similar like oh I'm going to talk to him and then I'm gonna steal everything we leave that in because we know players like it even if it's obviously is an exploit ok so so it's literally by design okay what about balance do you think is kind of harmful to balance we are not really focused on the main maxing we think about a game as something you were in which you should have fun especially if you're praying in cooperative multiplayer which is where most of these cases appear you just want to play like you are playing Dungeons &

Dragons you know you're playing a pen and paper session you're goofing around you're having a lot of fun there's reasons to continue in the world and that I think is the most important bit about RPG design okay thank you you're welcome here yeah sorry sorry yeah hello I'm just curious how you balance the crafting connected to the combat because there is clearly a connection between the two for the arrows bombs and this kind of shape so our crafting was one of the things that we weren't very happy about and so we think that we could have done a much better job on the crafting so this is something it's a topic on its own it came later it wasn't really embedded in the full design of the game and you feel it when you play the game also so we try to do that better for definitive edition but to be fair crafting is a real big problem when it comes to balancing and because those people don't like crafting a lot of people like crafting and it's hard to make it to mix them so if it were up to me I wouldn't have well it

was up to me so but I would have to think very hard again if I would put crafting into the game and because then that solves the problem obviously okay thank you you're welcome you talked a bit about how you flattened the AP system but it seems like the new system would be harder to balance just because you have fewer action points to use granularity could you talk more about why it was important to change that yeah it's so we noticed that people didn't care about the action points anymore when you had like eight or nine action points because obviously it was easy to scale but it just you didn't feel that there was no sense that it was important with two or three action points what we saw is people were arguing about it I'm gonna do this and then I'm gonna do this and that's literally what we wanted to promote so you're right I mean that just daggers on its own was already a problem right what are you gonna do one daggers one action point the other one is two action points or not but then what do you do with weapons that take two action points so those kinds of discussions but before that was more important that people would reason about our action points and

we took those problems with it we knew that that was going to be a problem gotcha Thanks right well we have time for one more question then we have to go to the wrap-up room according to my notes also you have to fill in your session evaluation things five stars please thank you so one last thing she was here before me you talked about reducing they're too many cooks situation I was wondering if you had any more ideas or things that you did that helped the whole studio get on board with it with the too many cooks well we explained it all right so we said like guys you have to leave us alone because otherwise we can't get our stuff done and there's this is this is a really good question because it's a hard problem right so on the one hand you don't want to be top-down heavy and you want to maximize what you get from the bottom-up on the other hand you still have to be able to work with it so I think this I don't have a ready solution for it I think it's a balance that you're gonna have to walk continuously and you're gonna have to adapt as you go and you'll make mistakes I mean because

it's which is fine making mistakes is fine in this and communicate a lot let's see the only solution I have I mean we're sudhi that went from 42 and we're now 200 people which has been a quite a ride but it's also complicated or life's tremendously because we're still some of our processes are still built for 40 people and that doesn't work with 200 people so we're discovering as we go right okay if you have any more questions we can talk about it and I don't know where the wrap up room is but there is a wrap up room and the wrap up room if somebody can show where the wrap up room is thank you alright [Applause]