Updated character heights comparison: tall female, medium male and tall male! (+ Statistical analysis)
WARNING: THIS POST IS EXTREMELY LONG, SO JUST GO TO THE END FOR THE FINAL ESTIMATES IF YOU DON'T WANT TO READ!
Oh boy...this is going to be a VERY long post, SORRY IN ADVANCE.
So, to start things off, as I'm sure you're all aware, there has been a lot of discussion over the years regarding the character heights in Genshin. Since there is no official material that gives us that information (which is stupid by the way) we have to resort to pixel measurements to create educated guesses about theses heights. today, I'll bring you some pixel measurements taken from in-game screenshots that allows us to compare the height ratios between different characters as well as my calculations based on those pixel measurements. Hope you enjoy!
NOTE 1: Here is something extremely important, while the pixel measurements MAY vary between the pictures, which you will see shortly, that doesn't mean that there is something with the statistics. This happens because moving the camera, having the characters be in a certain pose or tilting their heads all influence on the pixel count for these calcs. That doesn't mean they are inherently flawed, it just means that there is a margin of error we have to account for. What TRULY matters is the ratio between the characters being maintaned. For example:
\- Let's say that we have a screenshot where character A measures at 100 pixels and character B at 50, meaning character B is half of character A's height;
\- Then we take another screenshot of them and notice that in this second one, character A measures at 110 pixels and character B at 55;
We might think we did something wrong and screwed up the measurements, but that is not the case! If we take a look we see that Character A grew 10 pixels from one screenshot to another. Given that they were originally 100 pixels tall, that means that they grew about 10% in size. Now, if we take a look at character B, we see that they grew, surprise surprise, 10% as well (they grew 5 pixels and 5 = 10% of 50)! This means that, although the pixel measurements for both of these characters are different, the ratio they express between each other and themselves is consistent, meaning we can conclude that difference came from the angle of the camera, character's idle, etc.
NOTE 2 (Spoilers for 6.3 AQ): "Can't you just put these models on Blender and see how they measure?" It's conplicated. For one, the way a character model measures in and of themselves is NOT proof of how tall they are. What do I mean? Well, the most recent example is Dottore. Up until 6.3, he used a model in the cutscenes he appeared, however, in 6.3's AQ, during the first boss battle that you have with him, he uses an updated boss model which is CONSIDERABLY taller than his base cutscene model. Mind you, not the model he turns into at the end of the AQ, but the one used in the fight where traveler gets kidnapped. Both of them are models for Dottore and both represent the same character with no change whatsoever, and yet, they measure wildly different in Blender. So I truly do believe that using pixel measurements among the characters works better than relying on Blender.
Proof of the Blender situation (Image 1): https://ibb.co/mCFn15DJ
Oooookay, so now we can FINALLY start with the measurements. Three characters were chosen for this, these being Flins, Durin and Beidou ("Why Beidou? YOU WILL SEE WHY). Flins and Durin represent the tall and medium male model respectively.
So, for Flins' and Durin's height comparisons, we have these four images here:
Image 2: https://ibb.co/hxBJBBH6 (Flins' pixel height, side by side comparison);
Image 3: https://ibb.co/Lhshn3V2 (Durin's pixel height, side by side comparison);
Image 4: https://ibb.co/0pgNpJb0 (Durin's pixel height, front to front comparison);
Image 5: https://ibb.co/SXfJGQMD (Flins' pixel height, front to front comparison).
NOTE 4: You see that there are two measurements, both L and A for the pixels. Just ignore the L, the A is the actual height and what matters.
As we can see, the pixel measurements stay relatively the same between the two sets of images (side to side and front to front) with minor increases and decreases in measurements.
if we use the first set (side by side) we see that the ratio between Flins' and Durin's heights are 533/629, which means that Durin is as tall as roughly ≈ 84.73% of Flins' height.
By using the second set, utilizing the same calcs for the ratio we obtain that Durin is as tall as roughly ≈ 84.22% of Flins' height
GOOD, that means that the ratio stayed relatively the same. Now we can make a central estimate by simply making a median of the data we got (just add up 84.73 + 84.22 and divide the result by 2) which leads us to an estimate that Durin is roughly 84.47% of Flins' height.
For our margin of error, we will use the half-range error instead of the standard deviation account, because our dataset is small with few samples. That just means that we will get the range of error (84.73 - 84.22, which is equal to 0,51) and will divide it by 2. This results in a margin of error that 0,255% higher or lower than our central estimate of 84.47%.
GREAT, RIGHT?! So, now that we have a good ratio between the tall and the medium male model we need something...or someone who we can compare it to. That's where Beidou enters.
A lot of people, when working with pixel measurements like to compare the character heights to in-game objects that exist in the real world. I don't find this method particularly good, because it involves extrapolating the measurements of objects we see in game based on their real world counterparts, which involves a lot of assumptions as well as depending on which object it is.
image 6: https://ibb.co/chc6hgf4 (Image of Ororon by a door in the Serenitea teapot. Normally, doors are very standard sized, around 2 meters tall in the real world, or 6'7ft tall roughly....but Ororon is literally half the size of this door, so if it was 6'7ft tall...Ororon would just be a little over 3ft tall?????)
Beidou is here because she provides another solution for us! There is an NPC in Liyue called tea master Liu Shu, who is a storyteller. In one of his stories, he mentions the time where captain of the Crux fleet (Beidou) slayed a beast of the sea. I know...what does this have to do with character heights? Well, in this story there is a particular description that Beidou stood 6ft tall proudly! This is one of the ONLY times we get an ACTUAL number for a character's height in the entire game.
Now, I HAVE to mention this: there is a lot of critiscim with taking this NPC's lines seriously and comparing heights. He is a storyteller and is prone to exaggerating descriptions to give his stories more flare and this is why most people like to do object-based comparisons instead of comparing models to Beidou. However, the reason why I used it is because, for better or worse, as I said, this is one of the only times where we get an actual number to rely on in the game, and even if it is exaggerated, given that Beidou is a relatively known person in Liyue, I'd reckon that her height IS in the ballpark of being 6ft tall, otherwise, it would be quite ridiculous to narrate a story like that based on a person who is like...5'5 or something. That is to say, Beidou is the base model which I'll compare Flins and Durin to, with the assumption of her being 6ft tall. But don't worry, I'll also put results if we assume she was 5'9ft tall at the end of the post :)
image 7: https://ibb.co/DH7PyKnW (The Wikipedia page with the dialogue of tea master Liu Shu where he mentions Beidou "standing over 6ft tall")
NOW LET'S GET BACK TO PIXEL MEASUREMENTS!
Our next set will be pixel measurements between Flins and Beidou. As I said, the pixels may differ from the first set of photos but the proportions stay the same within this set. One thing I'd like to mention, though, which will repeat in Durin's set later is that, there is a larger gap between Beidou's pixel height in the side to side screenshot and the front to front screenshot. If you look, you'll notice she is looking directly at the camera with her head tilted and, believe it or not, that makes a really big difference in the pixel count, not so much in the ratio, but I'm saying to be transparent and so you understand why there is such a difference in the first place. Here are the images:
Image 8: https://ibb.co/FkSDp0bN (Flins' pixel height, side to side comparison);
Image 9: https://ibb.co/WNkxRVcP (Beidou's pixel height, side to side conparison);
Image 10: https://ibb.co/G3RZxS2n (Flins' pixel height, front to front comparison);
Image 11: https://ibb.co/hxH6g8LH (Beidou's pixel height, front to front comparison)
Let's do the same thing we did on the Flins/Durin sets with these ones. So, on the side to side sets, we see that Flins' height is 740 pixels and Beidou's is 672. We make the ratio 672/740, which lends us to a proportion of 90,81%. Meaning that Beidou has 90,81% of Flins' height.
If we take the second set, as I've said, we can see a noticeable difference in Beidou's measurements where she went from 672 pixels to 653, while Flins stayed relatively the same, going up just to 742. Doing this ratio we lend on exactly 88%, meaning that Beidou here corresponds to 88% of Flins' height.
HOWEVER, because of the inherent bias here in our dataset, we DO NOT average out the range between the ratios we've obtained like we did with Durin's and Flins' in the first sets. Instead we set a confidence interval between our upper bound limit (which is 90,8%) and our lower bound limit (which is 88%). The difference between these two numbers is an interval of 2.8%. Therefore, we can pick a value in the middle, preferentially and extrapolate this interval. This is all to say that our estimate of Beidou's height would be 89.4% of Flins' height, with a margin of error of 2,8%.
Next, we finally have the final set of screenshots that are pixel measurements for Durin and Beidou. I know what you might say: "You already know Durin's proportion in comparison to Flins and you know Flins' proportion compared to Beidou, which you know the height of, so can't you just calc Durin's height already?" Yes...yes I can, but by comparing Durin with Beidou we can also amplify our dataset and end up with more information on the accuracy of our calculations up until now, so it is worth the effort! Here are the images as always:
Image 12: https://ibb.co/Wp6qMq4G (Beidou's pixel height, side by side comparison);
Image 13: https://ibb.co/Fkgd1KFs (Durin's pixel height, side by side comparison);
Image 14: https://ibb.co/b5KgYvCT (Durin's pixel height, front to front comparison);
Image 15: https://ibb.co/xt9Q3pHj (Beidou's pixel height, front to front comparison)
Funnily enough, we see the same problem that occured with Beidou and Flins' comparisons happening here as well, by basically the same measurements. Which, in and off of itself is good, because at least it show consistency.
By running the same calcs we did up until now on the first set of side by side comparisons, we get that the ratio is roughly 96,12% from Durin, which means his height corresponds to that percentage of Beidou's height.
Doing the same thing on the second set of front to front pics we get a ratio of 96,81% for Durin. Now, although this set of information is biased because of Beidou's change in measurements, we see that the interval is a lot closer than what it was between the ratios of Flins' and Beidou's comparisons, with less than a percentual difference between the two. This allows us to average out the numbers again which will give us 96,46% as our central estimate as to how tall Durin is compared to Beidou, and to estimate our margin of error, we can simply do the half range error again, which gives us a margin of error of 0,345% that could be lower or higher than our central estimate.
NOW, with ALL of that said, we can finally, FINALLY create our estimates for the heights using all of these ratios. First, let's do a direct comparison with Beidou.
Beidou (6ft - 183cm) is 89.4% of Flins height. Which means Flins is 204.7 cm tall or 6ft 8.5 in.
Durin is 96,46% as tall as Beidou (6ft - 183 cm ), which means he is 176.5 cm tall, or 5ft 9.5 in tall.
We can also do an indirect estimate. Kind of like this: since I know Beidou's height and the ratios of both her with Durin and her with Flins, how does that compare to our previous ratio of Durin with Flins?
This is extremely important, because remember, our results for the Flins/Beidou ratio had a lot bias and our confidence interval was much higher than the other estimates. To ensure a better estimate of Flins' height, we can use a ratio with lower bias, in this case, Durin's and Beidou's ratio which determined that Durin was 176,5 cm. Then, we can plug this number in our direct comparison ratio between Flins and Durin:
Durin (5ft9.5in - 176.5 cm) is 84.47% as tall as Flins, which makes Flins 208.9 cm in size, or 6ft 10.2 in.
As incredible as it seems, this is a better estimate than the one above, even though it is bigger, because the bias in that comparison was a lot lower than with Beidou and Flins.
As promised, if we assume Beidou is 5'9 as her Blender model suggests and other height charts say:
Durin will be approximately 168.8 cm/5ft 6.5in
Flins, using the biased Beidou/Flins ratio is 195.8 cm/6ft 5in
Flins using the less biased Flins/Durin ratio is 199.9 cm/ 6ft 6.7 in.
THERE, DONEEEE. Hope this was helpful and I hope all of you have a great day <3
Oh boy...this is going to be a VERY long post, SORRY IN ADVANCE.
So, to start things off, as I'm sure you're all aware, there has been a lot of discussion over the years regarding the character heights in Genshin. Since there is no official material that gives us that information (which is stupid by the way) we have to resort to pixel measurements to create educated guesses about theses heights. today, I'll bring you some pixel measurements taken from in-game screenshots that allows us to compare the height ratios between different characters as well as my calculations based on those pixel measurements. Hope you enjoy!
NOTE 1: Here is something extremely important, while the pixel measurements MAY vary between the pictures, which you will see shortly, that doesn't mean that there is something with the statistics. This happens because moving the camera, having the characters be in a certain pose or tilting their heads all influence on the pixel count for these calcs. That doesn't mean they are inherently flawed, it just means that there is a margin of error we have to account for. What TRULY matters is the ratio between the characters being maintaned. For example:
\- Let's say that we have a screenshot where character A measures at 100 pixels and character B at 50, meaning character B is half of character A's height;
\- Then we take another screenshot of them and notice that in this second one, character A measures at 110 pixels and character B at 55;
We might think we did something wrong and screwed up the measurements, but that is not the case! If we take a look we see that Character A grew 10 pixels from one screenshot to another. Given that they were originally 100 pixels tall, that means that they grew about 10% in size. Now, if we take a look at character B, we see that they grew, surprise surprise, 10% as well (they grew 5 pixels and 5 = 10% of 50)! This means that, although the pixel measurements for both of these characters are different, the ratio they express between each other and themselves is consistent, meaning we can conclude that difference came from the angle of the camera, character's idle, etc.
NOTE 2 (Spoilers for 6.3 AQ): "Can't you just put these models on Blender and see how they measure?" It's conplicated. For one, the way a character model measures in and of themselves is NOT proof of how tall they are. What do I mean? Well, the most recent example is Dottore. Up until 6.3, he used a model in the cutscenes he appeared, however, in 6.3's AQ, during the first boss battle that you have with him, he uses an updated boss model which is CONSIDERABLY taller than his base cutscene model. Mind you, not the model he turns into at the end of the AQ, but the one used in the fight where traveler gets kidnapped. Both of them are models for Dottore and both represent the same character with no change whatsoever, and yet, they measure wildly different in Blender. So I truly do believe that using pixel measurements among the characters works better than relying on Blender.
Proof of the Blender situation (Image 1): https://ibb.co/mCFn15DJ
Oooookay, so now we can FINALLY start with the measurements. Three characters were chosen for this, these being Flins, Durin and Beidou ("Why Beidou? YOU WILL SEE WHY). Flins and Durin represent the tall and medium male model respectively.
So, for Flins' and Durin's height comparisons, we have these four images here:
Image 2: https://ibb.co/hxBJBBH6 (Flins' pixel height, side by side comparison);
Image 3: https://ibb.co/Lhshn3V2 (Durin's pixel height, side by side comparison);
Image 4: https://ibb.co/0pgNpJb0 (Durin's pixel height, front to front comparison);
Image 5: https://ibb.co/SXfJGQMD (Flins' pixel height, front to front comparison).
NOTE 4: You see that there are two measurements, both L and A for the pixels. Just ignore the L, the A is the actual height and what matters.
As we can see, the pixel measurements stay relatively the same between the two sets of images (side to side and front to front) with minor increases and decreases in measurements.
if we use the first set (side by side) we see that the ratio between Flins' and Durin's heights are 533/629, which means that Durin is as tall as roughly ≈ 84.73% of Flins' height.
By using the second set, utilizing the same calcs for the ratio we obtain that Durin is as tall as roughly ≈ 84.22% of Flins' height
GOOD, that means that the ratio stayed relatively the same. Now we can make a central estimate by simply making a median of the data we got (just add up 84.73 + 84.22 and divide the result by 2) which leads us to an estimate that Durin is roughly 84.47% of Flins' height.
For our margin of error, we will use the half-range error instead of the standard deviation account, because our dataset is small with few samples. That just means that we will get the range of error (84.73 - 84.22, which is equal to 0,51) and will divide it by 2. This results in a margin of error that 0,255% higher or lower than our central estimate of 84.47%.
GREAT, RIGHT?! So, now that we have a good ratio between the tall and the medium male model we need something...or someone who we can compare it to. That's where Beidou enters.
A lot of people, when working with pixel measurements like to compare the character heights to in-game objects that exist in the real world. I don't find this method particularly good, because it involves extrapolating the measurements of objects we see in game based on their real world counterparts, which involves a lot of assumptions as well as depending on which object it is.
image 6: https://ibb.co/chc6hgf4 (Image of Ororon by a door in the Serenitea teapot. Normally, doors are very standard sized, around 2 meters tall in the real world, or 6'7ft tall roughly....but Ororon is literally half the size of this door, so if it was 6'7ft tall...Ororon would just be a little over 3ft tall?????)
Beidou is here because she provides another solution for us! There is an NPC in Liyue called tea master Liu Shu, who is a storyteller. In one of his stories, he mentions the time where captain of the Crux fleet (Beidou) slayed a beast of the sea. I know...what does this have to do with character heights? Well, in this story there is a particular description that Beidou stood 6ft tall proudly! This is one of the ONLY times we get an ACTUAL number for a character's height in the entire game.
Now, I HAVE to mention this: there is a lot of critiscim with taking this NPC's lines seriously and comparing heights. He is a storyteller and is prone to exaggerating descriptions to give his stories more flare and this is why most people like to do object-based comparisons instead of comparing models to Beidou. However, the reason why I used it is because, for better or worse, as I said, this is one of the only times where we get an actual number to rely on in the game, and even if it is exaggerated, given that Beidou is a relatively known person in Liyue, I'd reckon that her height IS in the ballpark of being 6ft tall, otherwise, it would be quite ridiculous to narrate a story like that based on a person who is like...5'5 or something. That is to say, Beidou is the base model which I'll compare Flins and Durin to, with the assumption of her being 6ft tall. But don't worry, I'll also put results if we assume she was 5'9ft tall at the end of the post :)
image 7: https://ibb.co/DH7PyKnW (The Wikipedia page with the dialogue of tea master Liu Shu where he mentions Beidou "standing over 6ft tall")
NOW LET'S GET BACK TO PIXEL MEASUREMENTS!
Our next set will be pixel measurements between Flins and Beidou. As I said, the pixels may differ from the first set of photos but the proportions stay the same within this set. One thing I'd like to mention, though, which will repeat in Durin's set later is that, there is a larger gap between Beidou's pixel height in the side to side screenshot and the front to front screenshot. If you look, you'll notice she is looking directly at the camera with her head tilted and, believe it or not, that makes a really big difference in the pixel count, not so much in the ratio, but I'm saying to be transparent and so you understand why there is such a difference in the first place. Here are the images:
Image 8: https://ibb.co/FkSDp0bN (Flins' pixel height, side to side comparison);
Image 9: https://ibb.co/WNkxRVcP (Beidou's pixel height, side to side conparison);
Image 10: https://ibb.co/G3RZxS2n (Flins' pixel height, front to front comparison);
Image 11: https://ibb.co/hxH6g8LH (Beidou's pixel height, front to front comparison)
Let's do the same thing we did on the Flins/Durin sets with these ones. So, on the side to side sets, we see that Flins' height is 740 pixels and Beidou's is 672. We make the ratio 672/740, which lends us to a proportion of 90,81%. Meaning that Beidou has 90,81% of Flins' height.
If we take the second set, as I've said, we can see a noticeable difference in Beidou's measurements where she went from 672 pixels to 653, while Flins stayed relatively the same, going up just to 742. Doing this ratio we lend on exactly 88%, meaning that Beidou here corresponds to 88% of Flins' height.
HOWEVER, because of the inherent bias here in our dataset, we DO NOT average out the range between the ratios we've obtained like we did with Durin's and Flins' in the first sets. Instead we set a confidence interval between our upper bound limit (which is 90,8%) and our lower bound limit (which is 88%). The difference between these two numbers is an interval of 2.8%. Therefore, we can pick a value in the middle, preferentially and extrapolate this interval. This is all to say that our estimate of Beidou's height would be 89.4% of Flins' height, with a margin of error of 2,8%.
Next, we finally have the final set of screenshots that are pixel measurements for Durin and Beidou. I know what you might say: "You already know Durin's proportion in comparison to Flins and you know Flins' proportion compared to Beidou, which you know the height of, so can't you just calc Durin's height already?" Yes...yes I can, but by comparing Durin with Beidou we can also amplify our dataset and end up with more information on the accuracy of our calculations up until now, so it is worth the effort! Here are the images as always:
Image 12: https://ibb.co/Wp6qMq4G (Beidou's pixel height, side by side comparison);
Image 13: https://ibb.co/Fkgd1KFs (Durin's pixel height, side by side comparison);
Image 14: https://ibb.co/b5KgYvCT (Durin's pixel height, front to front comparison);
Image 15: https://ibb.co/xt9Q3pHj (Beidou's pixel height, front to front comparison)
Funnily enough, we see the same problem that occured with Beidou and Flins' comparisons happening here as well, by basically the same measurements. Which, in and off of itself is good, because at least it show consistency.
By running the same calcs we did up until now on the first set of side by side comparisons, we get that the ratio is roughly 96,12% from Durin, which means his height corresponds to that percentage of Beidou's height.
Doing the same thing on the second set of front to front pics we get a ratio of 96,81% for Durin. Now, although this set of information is biased because of Beidou's change in measurements, we see that the interval is a lot closer than what it was between the ratios of Flins' and Beidou's comparisons, with less than a percentual difference between the two. This allows us to average out the numbers again which will give us 96,46% as our central estimate as to how tall Durin is compared to Beidou, and to estimate our margin of error, we can simply do the half range error again, which gives us a margin of error of 0,345% that could be lower or higher than our central estimate.
NOW, with ALL of that said, we can finally, FINALLY create our estimates for the heights using all of these ratios. First, let's do a direct comparison with Beidou.
Beidou (6ft - 183cm) is 89.4% of Flins height. Which means Flins is 204.7 cm tall or 6ft 8.5 in.
Durin is 96,46% as tall as Beidou (6ft - 183 cm ), which means he is 176.5 cm tall, or 5ft 9.5 in tall.
We can also do an indirect estimate. Kind of like this: since I know Beidou's height and the ratios of both her with Durin and her with Flins, how does that compare to our previous ratio of Durin with Flins?
This is extremely important, because remember, our results for the Flins/Beidou ratio had a lot bias and our confidence interval was much higher than the other estimates. To ensure a better estimate of Flins' height, we can use a ratio with lower bias, in this case, Durin's and Beidou's ratio which determined that Durin was 176,5 cm. Then, we can plug this number in our direct comparison ratio between Flins and Durin:
Durin (5ft9.5in - 176.5 cm) is 84.47% as tall as Flins, which makes Flins 208.9 cm in size, or 6ft 10.2 in.
As incredible as it seems, this is a better estimate than the one above, even though it is bigger, because the bias in that comparison was a lot lower than with Beidou and Flins.
As promised, if we assume Beidou is 5'9 as her Blender model suggests and other height charts say:
Durin will be approximately 168.8 cm/5ft 6.5in
Flins, using the biased Beidou/Flins ratio is 195.8 cm/6ft 5in
Flins using the less biased Flins/Durin ratio is 199.9 cm/ 6ft 6.7 in.
THERE, DONEEEE. Hope this was helpful and I hope all of you have a great day <3
Commentaires