Daz+Maya研究報告(4)/ Daz+Maya research report (4) (Pixiv Fanbox)

那麼…我們終於完成了通往英靈殿的第一個儲存點，該來面對所遇到的問題了…

1—算圖速度

Arnold可以渲染出非常真實的漂亮光影，代價就是時間…

越複雜的光影效果，造成的雜訊就越多，就需要更高的AA sample參數，HDRI背景環境光可以輕易做出非常具真實感的光影，但造成的雜訊也是史詩級的…

目前我認為最理想的光影配置是：

HDRI sky dome*1 (低亮度，不產生陰影)

Direct light*2 (補光以及冷色系反光，不產生陰影)

Area light*1 (主力光線，產生陰影，綁攝影機正前方)

如此只有Area light會造成較大的算圖負擔，一張1500*1500左右的圖平均只要3分鐘左右即可算完，在3D製圖的領域來說算是很理想的速度了。

但是如果是用來做漫畫，仍然顯得有點笨重…目前我的作法是安排好鏡頭開始算圖後，先存一張低品質的圖丟進Photoshop裡，一邊編輯文字和特效一邊等待算圖完成，之後再匯入高畫質版取代，用這方法來勉強避開算圖造成的時間成本，但漫畫經常需要為了排版等需求重新編排，這時候要重算圖就會很痛苦，無論如何，算圖時間太長的話還是不太適合做漫畫。

2—編輯畫面與算圖結果的差異性

Arnold的各種特立獨行的材質球，導致很多效果無法在Maya hardware2.0 viewport上呈現，變成得要算圖出來之後才會知道長怎樣，算圖時間又很久，非常不利於測試，有時候圖都跑完了，才發覺有地方出問題或不好看，只能硬著頭皮再重跑一次…實在不是很靈敏。

3—編輯環境的LAG與BUG

Maya是個複雜的軟體，用博大精深來形容一點都不為過，相對的就是各種Bug與效能負擔…

在製作試作漫畫「夏天的結束」時，簡單的海灘場景加上三個人物，就讓我在編輯時各種卡頓了…這有點糟糕，因為這一場景+三人物已經算是最低限度的配置，還得再花點時間找出我的電腦能負擔的極限。

綜上所述，目前想到的解決方案有—

1—轉換製圖風格，單圖+文字，以文字為主圖片為輔，遊戲CG的形式

圖量越少，算圖時所遇到的困難就顯得比較無所謂，也比較能專注處理單一圖像裡的小細節

2—降低畫質，捨棄Arnold，使用Maya hardware2.0進行算圖

懂3D的人看到這條可能會有中風的衝動…但這其實解決很多問題，首先是所見及所得，能在編輯時就確定最終的樣子；算圖時間為0，同場景可以零成本拍出大量的角度；不使用Arnold場景相關處理，LAG的問題也會改善不少。

至於畫質…只能說比不上Arnold是一定的，但是以漫畫來說，其實畫質本來就不是說很重要吧？更關鍵的是人物動作，排版等等…

3—嘗試別的3D軟體，像是Blender

Blender的eevee render引擎能做到所見及所得的即時運算效果，品質似乎也很高，目前看來很符合我的需求，但換個3D軟體得花上很多時間重新熟悉快捷鍵等等操作，另外還有幾個功能上的考量不知道Blender是否能做到，因為Blender我完全沒用過，研究起來大概比Maya還要更花時間。

目前我打算三種解決方案都用，首先是先嘗試[3]Blender，如果能解決全部問題的話最好，不行的話就試試[2]Maya降低畫質的操作，也偶爾換換口味做點[1]文配圖的作品。

另外就是我也想嘗試製作Toon shader動漫風格的角色，目前已經稍微有點概念，但還沒有實際操作過，希望不要遇上太多技術障礙…

就這樣，我要開始狂刷Blender教學影片啦，請多指教:^)！

---------------------------------------------------------------------------------------------------------

So... we finally reached the first check point on the way to Valhalla, it's time to face the problems we encountered...

1—Render speed

Arnold can render very real and beautiful light and shadow, but the cost is time...

The more complex the light and shadow effect, the more noise it causes, and the higher the AA sample is required. The HDRI dome light can easily make a very realistic light and shadow, but the noise caused is also epic...

At present, I think the most ideal light and shadow configuration is:

HDRI sky dome*1 (low brightness, no shadows)

Direct light*2 (fill light and cool color reflection, no shadows)

Area light*1 (The main light, cast shadows, tied to the front of the camera)

In this way, only Area light will cause render burden. A 1500*1500 picture can be render in about 3 minutes on average, which is an ideal speed in the field of 3D graphics.

But if it is used to make comics, it still a bit slow... My current method is to arrange the camera and start the render, first save a low-quality image and throw it into Photoshop, edit the text and effects while waiting for the render to be completed. Then import the high-quality version to replace it. Using this method to barely avoid the time cost caused by the rendering, but the comics often need to be re-arranged for typesetting and other needs. At this time, it will be very painful to rerender. In any case, If the render time is too long, it is still not suitable for comics.

2—The difference between the editing screen and the calculation result

Arnold’s various maverick shaders cause many effects to be unavailable on the Maya hardware2.0 viewport, which is very unfavorable for testing. After I finished running, I realized that there was something wrong or not look good, I had to bite the bullet and run again... it was really not very flexble.

3—LAG and BUG of editing environment

Maya is a complex software. epic I'll say. The opposite is all kinds of bugs and performance burdens...

When making the testing manga "Summer's End", the simple beach scene and three characters made me lag when editing... This is bad, because one scene + three characters is already at the minimum requirement. I have to spend some more time figuring out the limits of what my computer can afford.

In summary, the solutions currently thought of are-

1—Convert the drawing style, single image + text, with text as the main, like game CG

The smaller the amount of images, the less pain of the difficulties encountered when render, and the better able to focus on small details in a single image.

2—Reduce the image quality, abandon Arnold, use Maya hardware2.0 for render

People who understand 3D may have a stroke when they see this article... But it actually solves many problems, first of all what you see is what you get, you can determine the final look when you edit; the calculation time is 0, the same scene can be shot many angle at zero cost; without using Arnold scene-related processing, LAG's problems will be improved a lot.

As for the picture quality...it can only be said that it is not comparable to Arnold, but in terms of comics, the picture quality is actually not that important, right? The more important thing is the character movement, typography, etc...

3—Try other 3D software, such as Blender

Blender's eevee render engine can achieve the real-time computing, and the quality seems to be very high. It seems to meet my needs at present, but it takes a lot of time to re-familiarize shortcut keys and other operations when changing a 3D software. There are also several functional considerations which I don’t know if Blender can do it, because I have not used Blender at all. It probably takes more time for research than Maya.

At present I plan to use all three solutions, first try [3] Blender, if it can solve all the problems, the best, if not, try [2] Maya reduce the quality of the operation, and occasionally change the taste[ 1] Works with text and pictures.

Also, I want to try to make Toon shader anime-style characters. I already have a little concept, but haven't actually operated it. Hope I won’t encounter too many technical obstacles...

That’s it, I’m going to watching a lot of Blender video tutorials, thanks for watching :^)!