Fan-Made Fix For Face Modification & Import Bugs Remaining After Patch 1.02
#801
Posté 13 avril 2012 - 10:45
#802
Posté 13 avril 2012 - 11:19
LolaLei wrote...
Well... she had all the same features as before, but she just didn't look the same.
Yep, common problem. The test really is simple: “does the player think the character looks the same”.
#803
Posté 13 avril 2012 - 11:25
I get the default female face instead of my customised one.
#804
Posté 13 avril 2012 - 11:30
Solid Gally wrote...
My import doesn't work AT ALL.
I get the default female face instead of my customised one.
Didja try deleting the face code from the save yet?
#805
Posté 13 avril 2012 - 11:38
For reference, here is the code: 733.J1P.118.C2T.78S.15S.31N.6D1.KGB.DL7.636.472

It was hilarious seeing the ME1 face emerge as I was adjusting the sliders. Definitely not a looker, but the ME1/2 faces look alike. I might have got a few ticks wrong, but overall they are very consistent. There were some minor differences, but I felt I was looking at the same 'person'. ME3 is very helpful in preventing people from making horrible modeling mistakes.
It was hard for me to get an in game shot in ME1 because the lighting on the Normandy is very dark. The one that I managed to grab just happens to add to the tragedy. It does express how I feel about this whole thing, though.

I did set the ME2 save to Complete and imported into ME3. The ME3 importer gave back the original ME3 face.
Modifié par Vhalkyrie, 14 avril 2012 - 12:06 .
#806
Posté 13 avril 2012 - 11:47
Vhalkyrie wrote...
Here's what happens when I take a ME3 facecode that looks ok and back port it all the way to ME1.
It was hilarious seeing the ME1 face emerge as I was adjusting the sliders. Definitely not a looker, but the ME1/2 faces look alike. I might have got a few ticks wrong, but overall they are very consistent. There were some minor differences, but I felt I was looking at the same 'person'..
Very interesting experiment. And yeah, it does look like the same person between 1 and 2. But 3 looks like someone else. Imagine that!
#807
Posté 13 avril 2012 - 11:50
Modifié par Vhalkyrie, 13 avril 2012 - 11:50 .
#808
Posté 14 avril 2012 - 12:20
#809
Posté 14 avril 2012 - 12:28
#810
Posté 14 avril 2012 - 01:36
Vhalkyrie wrote...
I did set the ME2 save to Complete and imported into ME3. The ME3 importer gave back the original ME3 face.
That's interesting. So does that mean for the people that didn't receive a preset face, it's not an import error but Bioware simply removed your original facial feature?
Uhh.. Jonah wrote...
Fight for the face.
I'm not sure why, but that made me burst out laughing
Modifié par Saiyan1126, 14 avril 2012 - 01:40 .
#811
Posté 14 avril 2012 - 01:48
Well, if Shepard can die, fall on a foreign planet, be put back together two years later and still have the same face in Mass Effect 2 then for pure continuity sake they need to fix this glitch for Mass Effect 3. So, if you won't fix it for the hoards of dissapointed fans, at least do it for the artists!
#812
Posté 14 avril 2012 - 01:52
Saiyan1126 wrote...
Vhalkyrie wrote...
I did set the ME2 save to Complete and imported into ME3. The ME3 importer gave back the original ME3 face.
That's interesting. So does that mean for the people that didn't receive a preset face, it's not an import error but Bioware simply removed your original facial feature?
It's more interesting than that. I made a copy of the save game and deleted the face code. That means ME3 should rebuild based on the LOD and face data in the ME2 save game, right? I got back the exact same ME3 face. They changed the mapping of the data points.
[edit] Mapping coordinates should be more accurate, but results are still off. Need more tests.
Modifié par Vhalkyrie, 14 avril 2012 - 02:17 .
#813
Posté 14 avril 2012 - 01:55
Saiyan1126 wrote...
Vhalkyrie wrote...
I did set the ME2 save to Complete and imported into ME3. The ME3 importer gave back the original ME3 face.
That's interesting. So does that mean for the people that didn't receive a preset face, it's not an import error but Bioware simply removed your original facial feature?
No, it simply confirms what we already knew. Slider positions don't mean the same thing in ME2 and ME3 (they have different variables assigned), so a character with exactly the same slider positions in ME2 and ME3 won't look the same, since slider 4 in ME2 may equal 1.789095 while the same slider in ME3 may equal 1.48234. But since Bioware decided to be lazy, they used slider positions to determine your face in the ME3 importer, unlike the ME2 importer, which used raw data (as proven by the fact that characters with hair styles not in the ME2 CC still imported fine).
PS: I'm guessing Biow-EA-re just did something similar to what the YAML workaround does with the ME1 save files, determine slider positions from the ME2 save file, create a face code, and input it into the ME3 CC
Modifié par Mingolo, 14 avril 2012 - 02:01 .
#814
Posté 14 avril 2012 - 02:05
Mingolo wrote...
PS: I'm guessing Biow-EA-re just did something similar to what the YAML workaround does with the ME1 save files, determine slider positions from the ME2 save file, create a face code, and input it into the ME3 CC
Yes, some others in the locked thread showed the results no better than the YAML method. You bring up a good point. I thought they translated the data points, but if they're just translating closest slider position, that makes sense why it doesn't actually match the coordinate mapping. Still digging into the raw data...I have found some interesting things, but still trying to gather more conclusive results before I report back.
Modifié par Vhalkyrie, 14 avril 2012 - 02:16 .
#815
Posté 14 avril 2012 - 02:24
Example:
Mass Effect 1's "Mouth Shape 5" gives the top lip a value of <6, 4, 160>.
Mass Effect 2's "Mouth Shape 5" gives the top lip a value of <6.5, 4.1, 161>.
Mass Effect 3's "Mouth Shape 5" gives the top lip a value of <6.3, 4.2, 160.75>.
When the face importer reads in value <6.5, 4.1, 161>, it uses the slider position equivalent in ME3 <6.3, 4.2, 160.75> by rounding to closest. For some facial features, this could occur by a larger margin. For some faces that are close, adjusting by a tick in either direction could make a closer match. In others, it won't make a difference because they widened/shortened the base feature by a noticable degree.
http://faceeffect.bl...you-put-it.html
ME2 gave us exact matches because they imported the raw data, which was why the "Accepted face" wasn't editable. If you wanted to change it, you had to go through custom face again and rebuild. In ME3, it doesn't import raw data. It uses the nearest slider positions.
Modifié par Vhalkyrie, 14 avril 2012 - 02:30 .
#816
Posté 14 avril 2012 - 02:27
Mingolo wrote...
No, it simply confirms what we already knew. Slider positions don't mean the same thing in ME2 and ME3 (they have different variables assigned), so a character with exactly the same slider positions in ME2 and ME3 won't look the same, since slider 4 in ME2 may equal 1.789095 while the same slider in ME3 may equal 1.48234. But since Bioware decided to be lazy, they used slider positions to determine your face in the ME3 importer, unlike the ME2 importer, which used raw data (as proven by the fact that characters with hair styles not in the ME2 CC still imported fine).
PS: I'm guessing Biow-EA-re just did something similar to what the YAML workaround does with the ME1 save files, determine slider positions from the ME2 save file, create a face code, and input it into the ME3 CC
That's what I meant. Looking back, I did kind of oversimplify it. Sorry, I've never discussed faces, or textures for that matter lol.
#817
Posté 14 avril 2012 - 02:32
Vhalkyrie wrote...
I found this and it confirms that I thought. They altered the geometry in such a way that slider positions will give off results.
Example:
Mass Effect 1's "Mouth Shape 5" gives the top lip a value of <6, 4, 160>.
Mass Effect 2's "Mouth Shape 5" gives the top lip a value of <6.5, 4.1, 161>.
Mass Effect 3's "Mouth Shape 5" gives the top lip a value of <6.3, 4.2, 160.75>.
When the face importer reads in value <6.5, 4.1, 161>, it uses the slider position equivalent in ME3 <6.3, 4.2, 160.75> by rounding to closest. For some facial features, this could occur by a larger margin. For some faces that are close, adjusting by a tick in either direction could make a closer match. In others, it won't make a difference because they widened/shortened the base feature by a noticable degree.
http://faceeffect.bl...you-put-it.html
ME2 gave us exact matches because they imported the raw data, which was why the "Accepted face" wasn't editable. If you wanted to change it, you had to go through custom face again and rebuild. In ME3, it doesn't import raw data. It uses the nearest slider positions.
So is there no way to alter these values like the people that used Gibbed to prevent clipping when they use Ashley's style of hair on their femsheps? I mean, I get what you're saying that some changing is just irrelevant, but any hope for those that aren't?
Modifié par ThinkIntegral, 14 avril 2012 - 02:32 .
#818
Posté 14 avril 2012 - 02:36
ThinkIntegral wrote..
So is there no way to alter these values like the people that used Gibbed to prevent clipping when they use Ashley's style of hair on their femsheps? I mean, I get what you're saying that some changing is just irrelevant, but any hope for those that aren't?
Altering the data values in Gibbed have given me very good results, but I don't want to get ahead of myself.
[In best Mordin voice] Need more tests. Come back later.
#819
Posté 14 avril 2012 - 02:39
Vhalkyrie wrote...
I found this and it confirms that I thought. They altered the geometry in such a way that slider positions will give off results.
Example:
Mass Effect 1's "Mouth Shape 5" gives the top lip a value of <6, 4, 160>.
Mass Effect 2's "Mouth Shape 5" gives the top lip a value of <6.5, 4.1, 161>.
Mass Effect 3's "Mouth Shape 5" gives the top lip a value of <6.3, 4.2, 160.75>.
When the face importer reads in value <6.5, 4.1, 161>, it uses the slider position equivalent in ME3 <6.3, 4.2, 160.75> by rounding to closest. For some facial features, this could occur by a larger margin. For some faces that are close, adjusting by a tick in either direction could make a closer match. In others, it won't make a difference because they widened/shortened the base feature by a noticable degree.
http://faceeffect.bl...you-put-it.html
ME2 gave us exact matches because they imported the raw data, which was why the "Accepted face" wasn't editable. If you wanted to change it, you had to go through custom face again and rebuild. In ME3, it doesn't import raw data. It uses the nearest slider positions.
Nice read. Do you think this is an example of cut corners, or do you think they had a legitimate reason for not importing the raw data? I noticed that faces are a lot more expressive than ME2 and I'm curious if that was the reason why they chose to import slider positions instead (I have no animation knowledge).
#820
Posté 14 avril 2012 - 03:00
Saiyan1126 wrote...
Nice read. Do you think this is an example of cut corners, or do you think they had a legitimate reason for not importing the raw data? I noticed that faces are a lot more expressive than ME2 and I'm curious if that was the reason why they chose to import slider positions instead (I have no animation knowledge).
The changes to the LOD (which means Level of Detail in computer graphics geometry, for those who don't know) does make the base mesh models more complex. Does this mean they couldn't import the raw data? No, it does not. If I can do it by hand, it can be done. What can be done is taking samples of key points in a structure, then extrapolating the gaps. This is not necessarily trivial (but I am making assumptions and erring on the side of complexity), and it would result essentially be the same as ME2 import - you could not further change the shape, eye, hair color/style, etc without going back through the custom appearance and rebuilding. They made a decision to force models to closest sliders so you could further modify your character. Some of these features on the sliders were purposefully 'softened' on femShep. Hawk noses and strong jawlines are harder to achieve. I'm sure they thought this was would be a more desirable change to most people, as I recall a lot of complaints that femShep wasn't feminine enough. I think they could have made softer features available, while retaining backward compatibility.
This is why many of us who used more extreme ends of the slider (like Clarian and myself) end up with radically different results. Others end up with more subtle degrees, but we can still tell something is off.
Modifié par Vhalkyrie, 14 avril 2012 - 03:26 .
#821
Posté 14 avril 2012 - 03:31
Phoenix did a lot of testing on it, the failure of the method has something to do with these variables used to determine LODs, which are what really determine face shape, and that they do not change when you change these variables, even if you delete the LODs and force the game to recreate them. The explanation was actually pretty complicated, I couldn't follow the whole thing. And direct copy/paste of the LODs does not work, because they are different in the two games. Woodbyte is the one currently working on how to import the LODs from ME2 to ME3.
Someone even made a long graph showing the face shape variables that you should change and the ones that you shouldn't, to get the best results.
Here are my results changing the raw face variables in Gibbed:
ME2

ME3 YAML workaround/patch + further tweaking

ME3 YAML workaround/patch + further tweaking + raw variables changed in Gibbed

As you can see, there's not much improvement from the YAML to the changing of raw variables in gibbed (the eye shape and the new annoying hair style keep being the biggest differences), except the eye color. Like I said, Phoenix said this had something to do with the LODs.
Modifié par Mingolo, 14 avril 2012 - 03:46 .
#822
Posté 14 avril 2012 - 03:37
OffsetBones has the most noticeable effect on structure. Vector and Scalar parameters alter the degree of certain variables, but mostly coloring and intensity.
Also, the reason LOD mapping won't work one to one is because there are more vertices. You can't translate the points without knowing where the lesser detail (in ME2) correspond to in ME3.
Modifié par Vhalkyrie, 14 avril 2012 - 03:47 .
#823
Posté 14 avril 2012 - 03:44
If I remember correctly, changing OffsetBones actually produced a worse looking character for Clarian.
#824
Posté 14 avril 2012 - 03:51
Also, translating some of the number directly hasn't always given a good result. It does on some features, not on others. Translating the eye position gave clipping issues, for example, so I scaled those back to default. Changing the brow height was possible, but not with the exact ME2 numbers. Anyway, I was trying not to report anything before I was positive about what I am seeing.
Modifié par Vhalkyrie, 14 avril 2012 - 03:56 .
#825
Posté 14 avril 2012 - 03:59
http://faceeffect.bl...-version-1.html
PS: it also has all the testing she did with more stuff and other things
Vhalkyrie wrote...
Also, translating some of the number directly hasn't
always given a good result. It does on some features, not on others.
Translating the eye position gave clipping issues, for example, so I
scaled those back to default. Changing the brow height was possible,
but not with the exact ME2 numbers. Anyway, I was trying not to report
anything before I was positive about what I am seeing.
Yeah, the graph that I mentioned pointed out exactly what variables you should change, and which you shouldn't (to avoid, for example, the clipping issues). Now, if I could only find it...
Modifié par Mingolo, 14 avril 2012 - 04:03 .




Ce sujet est fermé
Retour en haut





