3D Realtime Character [WIP]
For which Engine - XNA and M3G Application
Current Polycount -
Max Polycount - 10.000 for XNA, 1000 for M3G
Texture Size - 1024 or 2048 for XNA and 512 for M3G
Description of background information on piece - Realtime Character for my diploma thesis
Where are you looking to receive criticism - Especially Modelling and Edgeloops
_________________________________________________
Forword:
At the moment i´am writting my diploma thesis about Realtime Character Development with an implementation in XNA and M3G for Java. I have less experience in Character building and this should change with this project. I hope some poeple of you could help me with this.
My concept is based on the appleseed characters, because i´am not a skilled painter or drawer. The main aspect here is the modelling of the character, so it would be well animated and has less polycount.
Concept:
Image Planes:
I blocked the basic structre of the model. It would be very helpful, if you could help me with the edgeloops of the model.
Especially for the loops on the front of the hips down the foot. At the back of the Character down the foot. And the Loops at the back from the arms to the back.
At the moment the parts for the upper body, lower body, Head, armorplate and boots are seperated.
Insert Wireframe
Best regards and thanks for your hopefully help.
Current Polycount -
Max Polycount - 10.000 for XNA, 1000 for M3G
Texture Size - 1024 or 2048 for XNA and 512 for M3G
Description of background information on piece - Realtime Character for my diploma thesis
Where are you looking to receive criticism - Especially Modelling and Edgeloops
_________________________________________________
Forword:
At the moment i´am writting my diploma thesis about Realtime Character Development with an implementation in XNA and M3G for Java. I have less experience in Character building and this should change with this project. I hope some poeple of you could help me with this.
My concept is based on the appleseed characters, because i´am not a skilled painter or drawer. The main aspect here is the modelling of the character, so it would be well animated and has less polycount.
Concept:
Image Planes:
I blocked the basic structre of the model. It would be very helpful, if you could help me with the edgeloops of the model.
Especially for the loops on the front of the hips down the foot. At the back of the Character down the foot. And the Loops at the back from the arms to the back.
At the moment the parts for the upper body, lower body, Head, armorplate and boots are seperated.
Insert Wireframe
Best regards and thanks for your hopefully help.
Replies
http://www.sideshowtoy.com/?page_id=4489&sku=6363
http://www.sideshowtoy.com/?page_id=4489&sku=6389
those should help you with the details.
Any suggestions about the edgeloops?
Could need some help with the loops at these regions and the biceps and triceps.
I made some deformation tests and the main problem area is at the moment between thigh and hip.
Any suggestions before i go further:
And now i used it, because lack of time. :-)
I admit that I do like the narrow waist in the first concept art, compared to more bulky front and back.
The Polycount at the moment is at about 9200 Tris.
I should also mention that char has not to be 100% anatomically correct. Mayabe I will correct the issue of the shorter legs after my diploma thesis, but at the moment i actually completed the rig because i tested the deformation with this.
At the moment i ´am thinking about how i could place my assets like ammopacks, holster etc... an how i could rig this on the final model ? Because i don´t know which bones should influence this additonal parts, so they move correct with the body.
The Translation of the description means: How to paint the weights of the underlaying vertices. Because the parts of the Object will be combined before the skinning process.
The duplicated pieces are overlapping at the moment, because they were just duplicated. I have to layout all of them too.
But that kneepad is real dense. Also,push the groin out abit
So if you go that route, I would save out one channel with uvs overlapping on mirrored parts and another with the overlapping parts aside. This way when you bake, you can quickly switch between the two, and don't get errors on your normals.
Lastly, just cram them all in there as tight as you can, but be carfull that they arn't too close, so they don't collide at all on you lower resolution maps. Put the most important peices like the face in first with the size you want and build the other clusters around that.
Nice work so far, I really like your approach and how you are taking the time to carfully plan things out. Most people make the mistakes first and ask questions later which certainly isn't the right way to do things. I look forward to seeing more!
Now i made for the xna - model 3 Maps. one for the assets including the armor pieces and the boots, one for the body including the shoulder plates and one for the head. It would be possible to include all in 2 texture maps, but i think there will be no problem in using 3 maps.
I plan this wisley because my timeplan is very dense. I have to wrote the whole documentation after this (at the moment 96 pages, but there will follow a lot of pages about the xna and m3g part). My dateline for all of this is the 15. november. And unfortunately this blocks me in doing a properly well designed and detailled model. Maybe i will rework this model after this project.
now i will start with the sculpting and detailling part in zbrush.
That said, the high poly mesh looks really nice, and there is no reason that you can't capture that same look in your lowply with the polygon count you have currently on the low poly face. Using a program like polyboost in 3ds max can make this really easy. There are others also, you can retopologize in zbrush 3 (but it kind of sucks in my opinion) and if you can get on the beta testing crew, a really awsome program called topogun. Even, simply pushing verts just outside the high poly within your 3d app, could help a great deal.
I know you said your time is limited, but I think if you took the steps to rebuild your low poly based soley on the high, you will get far more realistic models.
Unfortunately the mainbody area esspecially the arms are also not so good for the details i wanted to apply.
The back seems to have to many poly's.I would get rid of those sharp edges on his nose,doesn't really look realistic IMO.
The shape of the eyes needs work too.
Keep it up.
Little Update on the upper body.
Either way keep it up.
Oh yeah, movento is right about the back, it can certainly be optimized more.
You might be making the polycount but they can probably help more with the sillohete if placed in other areas.
I will try get the chest fixed, but i dont know if this possible. About the polycount in the back...i dont know, if could delete those edges without getting issues on my UVs, but i will try this afterwards. Unfortunately The heavy polycount on the back is a result of the edgeloops on the front.
But first some hours sleep:
Ok, here the basic color + ao - pass and normal map. Any help / hint / crit. would be welcome:
As for the texture, I would say just keep working at the outfit it is looking nice so far, but for the face it's lacking a lot of natural skin tones such as yellows and reds and imperfections with the skin. Also the hairline and the eyebrows look too clean.
I updated the Head Texture.
Head Update:
Look at your proportions, I think that's where things changed.
Also since you have nice UV, bake in a layer of Ambient Occlusion, tone down the color saturation of your texture maps and modulate them with the AO
Thx for your help. Unfortunately at this stage it isn´t possible to change any major proportions and placements, because i started rigging.
What i can do is changing textures. And i truly added the ambient occlussion to all of my parts. like here:
i am already in the weighting progress:
Rigging-Test
Walk Cycle
I run into a major problem. I exported my skinned model as a fbx and tried to import it in my xna application, but i get this error:
Fehler 1 Building content threw ArgumentOutOfRangeException: Das angegebene Argument liegt au
You won't be able to say that if an Art Director or higher ups wanted changes. Gotta learn to start scrapping things that aren't working. Never get attached to your creations
~t
I never modeled , textured, rigged and animated a complete char. Also i have to code a xna application AND a m3g application. Beside this, the writting of the thesis. At the moment i have 110 pages and this is about 1/3 of all which i have to write.
Just 7 weeks remaining and a lot of stuff to do. I can Fix all these problems after i had done this project. For my thesis, a simple prototyp of a char would be enough. I think this stage is much more than a prototyp and enough work for thesis. Beside this fact, unfortunately the main focus on the grade for the thesis lies in the theoretical part and the writing of the thesis.
flip the uv of the wrong parts. bake and see if it helps
//Deklaration und Initialisierung #define MaxBones 59 //setzt die Maximale Anzahl der verwendeten Bones fest float4x4 Bones[MaxBones]; float4x4 View; float4x4 Projection; float3 lightDir1 = float3(1,1,1); float3 lightDir2 = float3(-1,-1,-1); //Vertexshader mit Tangent-Space Berechnung mit Rückgabewert output VS_OUTPUT VSNormals(VS_INPUT input) { VS_OUTPUT output; float4x4 skinTransform = 0; skinTransform += Bones[input.BoneIndices.x] * input.BoneWeights.x; skinTransform += Bones[input.BoneIndices.y] * input.BoneWeights.y; skinTransform += Bones[input.BoneIndices.z] * input.BoneWeights.z; skinTransform += Bones[input.BoneIndices.w] * input.BoneWeights.w; float4 pos = mul(input.Position, skinTransform); pos = mul(pos,View); pos = mul(pos,Projection); float3x3 tgtSpace; tgtSpace[0] = mul(input.Tangent, skinTransform); tgtSpace[1] = mul(cross(input.Tangent, input.Normal),skinTransform); tgtSpace[2] = mul(input.Normal, skinTransform); float3 tgtLightDir1 = mul(lightDir1, tgtSpace); float3 tgtLightDir2 = mul(tgtSpace, lightDir2); output.Position = pos; output.TexCoord = input.TexCoord; output.Data1 = normalize(tgtLightDir1); output.Data2 = normalize(tgtLightDir2); return output; } //Basis Normalmap - Shader float4 PSNormals(VS_OUTPUT input) : COLOR0 { float4 outColor = float4(1,1,1,1); float3 normal = tex2D(N_Sampler, input.TexCoord); float diffuse = saturate(dot(normal, input.Data1)) + saturate(dot(normal, input.Data2)); return outColor * diffuse; }when i change the code to this://Basis Normalmap - Shader float4 PSNormals(VS_OUTPUT input) : COLOR0 { float4 outColor = float4(1,1,1,1); [COLOR=Red][B]float3 normal = 2* tex2D(N_Sampler, input.TexCoord) -1.0;[/B][/COLOR] float diffuse = saturate(dot(normal, input.Data1)) + saturate(dot(normal, input.Data2)); return outColor * diffuse; }i get this render:When i map just my normalmap texture to the model everything is facing correct .
In Maya the Normals are all facing the in the same direction, but i rotated some pieces in the normalmap. like you can see here:
http://www.d-mack.de/tmp/tmpPics/thesisWIP/XNA_NormalProblem01c.jpg[/img]
I´am new to xna and hlsl shading programming. I think the calculation of the normals results in the light direction and the colors in the normal map.
The colors in the normal map facing the right direction like you can see here:
i would think the problem lies in the calculation in the tangent space normals and the light direction which will be done here:
float3x3 tgtSpace;
tgtSpace[0] = mul(input.Tangent, skinTransform);
tgtSpace[1] = mul(cross(input.Tangent, input.Normal),skinTransform);
tgtSpace[2] = mul(input.Normal, skinTransform);
float3 tgtLightDir1 = mul(lightDir1, tgtSpace);
float3 tgtLightDir2 = mul(tgtSpace, lightDir2);
output.Position = pos;
output.TexCoord = input.TexCoord;
output.Data1 = normalize(tgtLightDir1);
output.Data2 = normalize(tgtLightDir2);
return output;
and here:
//Basis Normalmap - Shader
float4 PSNormals(VS_OUTPUT input) : COLOR0
{
float4 outColor = float4(1,1,1,1);
float3 normal = 2* tex2D(N_Sampler, input.TexCoord) -1.0;
float diffuse = saturate(dot(normal, input.Data1)) +
saturate(dot(normal, input.Data2));
return outColor * diffuse;
}
before i layout out my uvs it looked so:
I flipped the uvs for the red parts because the were duplicates of the right half of the character. And all pieces were blue before i layed out my uvs within the 0 and 1 coordinates. So i think, this is not the problem. (Correct me if i am wrong). But after i correct all, all uvs facing the same direction and were shaded blue.
if i compare to this pic
this could not be the problem.
Heres the layout of the final char with the uvs and the w- component all facing the same direction:
I'm pretty sure I could help you with flipping the uv's if it was 3ds max, in maya I have no clue. Someone here on polycount has an answer though I'm sure.
when its right in maya (but its openGL) and right in fx composer (with directX), there should be a way to correct this within my shader.fx file.
But unfortunately i very new to shader programming and have no clue where the fault in calculation for my normal map lies.
In the XNA Forum i got this hint for solving the problem but because i just started with HLSL Shader Programming i don´t have the clue how to solve this:
Here´s my complete shader.fx code:
//SkinnedModel HLSL - Shader Datei //Deklaration und Initialisierung #define MaxBones 59 //setzt die Maximale Anzahl der verwendeten Bones fest float4x4 Bones[MaxBones]; float4x4 View; float4x4 Projection; float3 lightDir1 = float3(1,1,1); float3 lightDir2 = float3(-1,-1,-1); //Deklariert die ColorTextur des Models texture Texture; sampler C_Sampler = sampler_state { Texture = <Texture>; MinFilter = Linear; MagFilter = Linear; MipFilter = Linear; }; //Deklariert die Normalmap-Textur des Models texture N_Texture; sampler N_Sampler = sampler_state { Texture = <N_Texture>; MinFilter = Linear; MagFilter = Linear; MipFilter = Linear; }; //struct-Definition für Vertexshader input struct VS_INPUT { float4 Position : POSITION0; float2 TexCoord : TEXCOORD0; float3 Normal : NORMAL0; float3 Tangent : TANGENT0; float4 BoneIndices : BLENDINDICES0; float4 BoneWeights : BLENDWEIGHT0; }; //struct-Definition für Vertexshader output struct VS_OUTPUT { float4 Position : POSITION0; float2 TexCoord : TEXCOORD0; float3 Data1: TEXCOORD1; float3 Data2: TEXCOORD2; }; //Basis Vertexshader mit Rückgabewert output VS_OUTPUT VSBasic(VS_INPUT input) { VS_OUTPUT output; float4x4 skinTransform = 0; skinTransform += Bones[input.BoneIndices.x] * input.BoneWeights.x; skinTransform += Bones[input.BoneIndices.y] * input.BoneWeights.y; skinTransform += Bones[input.BoneIndices.z] * input.BoneWeights.z; skinTransform += Bones[input.BoneIndices.w] * input.BoneWeights.w; float4 pos = mul(input.Position, skinTransform); pos = mul(pos,View); pos = mul(pos,Projection); float3 nml = mul(input.Normal, skinTransform); nml = normalize(nml); output.Position = pos; output.TexCoord = input.TexCoord; output.Data1 = nml; output.Data2 = 0; return output; } //Vertexshader mit Tangent-Space Berechnung mit Rückgabewert output VS_OUTPUT VSNormals(VS_INPUT input) { VS_OUTPUT output; float4x4 skinTransform = 0; skinTransform += Bones[input.BoneIndices.x] * input.BoneWeights.x; skinTransform += Bones[input.BoneIndices.y] * input.BoneWeights.y; skinTransform += Bones[input.BoneIndices.z] * input.BoneWeights.z; skinTransform += Bones[input.BoneIndices.w] * input.BoneWeights.w; float4 pos = mul(input.Position, skinTransform); pos = mul(pos,View); pos = mul(pos,Projection); float3x3 tgtSpace; tgtSpace[0] = mul(input.Tangent, skinTransform); tgtSpace[1] = mul(cross(input.Tangent, input.Normal),skinTransform); tgtSpace[2] = mul(input.Normal, skinTransform); float3 tgtLightDir1 = mul(tgtSpace, lightDir1); float3 tgtLightDir2 = mul(tgtSpace, lightDir2); output.Position = pos; output.TexCoord = input.TexCoord; output.Data1 = -2 * normalize(tgtLightDir1) +1; output.Data2 = normalize(tgtLightDir2); return output; } //Basis Pixelshader mit Textur float4 PSBasic(VS_OUTPUT input) : COLOR0 { float4 outColor = tex2D(C_Sampler, input.TexCoord); return outColor; } //Diffuse Pixelshader mit 2 Lichtern float4 PSDiffuse(VS_OUTPUT input) : COLOR0 { float4 outColor = tex2D(C_Sampler, input.TexCoord); float diffuse = saturate(dot(input.Data1, normalize(lightDir1)))+ saturate(dot(input.Data1, normalize(lightDir2))); return outColor * diffuse; } //Basis Normalmap - Shader float4 PSNormals(VS_OUTPUT input) : COLOR0 { float4 outColor = float4 (1,1,1,1); float3 normal = tex2D(N_Sampler, input.TexCoord) ; float diffuse = saturate(dot(normal, input.Data1)) + saturate(dot(normal, input.Data2)); return outColor * (diffuse); } //Basis Normalmap - Shader als einfache Textur float4 PSNormalsTexture(VS_OUTPUT input) : COLOR0 { float4 outColor = tex2D(N_Sampler, input.TexCoord); return outColor; } //Definition der verwendeten Techniken. //Grundtechnik mit Pixelshader und Vertexshader Version 1.1 technique SkinnedModelTechnique { pass SkinnedModelPass { VertexShader = compile vs_1_1 VSBasic(); PixelShader = compile ps_1_1 PSBasic(); } } //DiffusTechnik mit Vertexshader 1.1 und Pixelshader 2.0 technique DiffuseMap { pass Pass1 { VertexShader = compile vs_1_1 VSBasic(); PixelShader = compile ps_2_0 PSDiffuse(); } } //Normalmap Technik mit Vertexshader 1.1 und Pixelshader 2.0 technique BumpAsTexture { pass Pass1 { VertexShader = compile vs_1_1 VSBasic(); PixelShader = compile ps_2_0 PSNormalsTexture(); } } //Normalmap Technik mit Vertexshader 1.1 und Pixelshader 2.0 technique Bump { pass Pass1 { VertexShader = compile vs_1_1 VSNormals(); PixelShader = compile ps_2_0 PSNormals(); } }But how could i calculate the Tangents, BiTangents and Normals correctly?
1.unskin your mesh.
2.Select the parts of the mesh you want to change the scale on wether its in one mesh or multiples.
3.combine them( this is temporary, it will not affect the uvs at all). 4.Selelct the vertxes in the now combined mesh to which the change is relevent.
5.go animation menu create deformers and lattuce with the desired number of control points.
6.once you achive your desired change in proportion select the mesh delete history and separate to get your armor pieces back to thir own meshes.
go ahead and use the
skeleton>edit smooth skin> move skinned joints tool to correct your rig for the new proportions.
hope this was at least a lil helpful, like i said ti will take a lil tweaking but it does work. the company im with now is disorganized and i have to alter skinned models all the time.
-gl