Post-release YAP Session

Hey everyone, this will be a post of me talking about the stuff I've been through during the development of chapter 2 and what I plan to do in the future. The post will last for a bit, kind of like my uni essays. But this one I enjoy writing, because I LOVE yapping.

Development Time

Currently, I believe the concern about Sort of Justice is the development time. It took around 1 year for Chapter 2 to release and that had definitely left a bad impression. I've explained why it took this much time so let's not bring it up again. Anyway, let's take a look at the work that goes into developing a chapter.

- Writing

I write the dialogues of each chapter after the previous chapter is released. A whole chapter of dialogue usually spans around 8k-10k words. Even if I already have the chapter plotted out in a scene board, it's still going to take me a while. Let's say... three weeks. And that's optimistic. I plan to keep the same amount of content in a chapter going forward, similar to chapter 2, so let's try to keep the writing time to the minimal. Even so, it doesn't account for the time required to fix writing inconsistency, which happens very often even when the chapter is plotted out. In chapter 2, I needed to completely removed 2 full scenes to maintain the flow and continuity (and to reduce development time). I guarantee that this will happen in almost every chapter from now on; not exactly on the level of needing to delete 2 whole scenes but enough script editing to occupy a small portion of the development time. Though, I have to say, if there's a need for optimisation or correction to the script, I'm willing to trade my time for it. Nothing implodes faster than an inconsistent sloppy piece of mess.

- Visuals

Rendering and editing images takes up a larger portion of the development time. Previously, during the development of chapter 1, I employed an extremely inefficient workflow that pretty much tripled the development time. During chapter 1, I would put the assets together in a scene, pose the characters and set the lights in one frame, render it, then move on to the next frame. The main reason why I would do this is to avoid unforeseen human error such as mesh-clipping, lighting mistakes, incorrect poses, and many more. This concern originated from my inexperience to optimise renders in a way so that each render doesn't take 15 minutes to finish. You tend to be very careful when it takes such a long time for just one render. But guess what? I can mess it up somehow.

Therefore, I decided in chapter 2, I'm giving it a bit of a change. Before rendering a scene, I would first pose the characters on all frames, then set the lights, then check for errors, then render all frames. I have also limited the render time to 5 minutes max, so regardless of the converged ratio, I can finish one render in less than 10 minutes. However, the 5 minute limit only applies to render with optimised scenes.

I should actually explain what is converged ratio first. Basically, converged ratio is a metric used to present how many pixels are rendered; in other words, it tells us how complete a render is. The converged ratio is directly affected by how surfaces reflect or absorb light in a scene. Depending on the environment, surfaces are either easier or harder to be hit by light sources, and hence, easier or harder to be rendered. Some indoor scenes take a long time to complete because of how many surfaces there are for the light to reflect on, and the converged ratio would still be around 30% after 5 minutes. Let me put this into a perspective. With my hardware, an outdoor scene with a large amount of props/characters will reach 92% converged ratio in 5 minutes. An indoor scene with sufficient lighting and a minimum amount of props/characters will reach 76~% in 5 minutes. An indoor scene with sufficient lighting but a large amount of prop/characters will reach 42% converged ratio in 5 minutes.

So, how does each respective converged ratio look like? At 85-99%, the image finished rendering with the most amount of detail and the least amount of noise. At 70-80%, the image finished rendering with an average amount of detail and noise. At 40-50%, the image finished rendering with little amount of detail and large amount of noise. Noisy images can be fixed by using a denoiser, but the outcome will be determined by the quality of the raw image. The action of denoising is basically removing the noise/unwanted grains in an image by filling them with coloured pixels through filtering and machine-learning. When used on a high converged ratio image, such as 80-99%, the details are mostly preserved. However, when denoising low converged ratio image, details are very poorly presented, because the details were never there to begin with. Low converged ratio/noisy images can hardly present an adequate level of detail; which is why if I want to preserve image quality, I need to render the image with a converged ratio of at least 55-70%. This often take 10-15 minutes in a heavy scene. Now, you know part of the reasons of my struggle.

- Audio

Sound effects and music are both major contributing factors in Sort of Justice. However, as I've expressed in one of the previous dev logs, it takes a huge chunk of development time to go over all the audio works. For me, I have to wait until both the writing and visuals to be done before working on the audio. I would go through the dialogues one by one and insert sound effects and music accordingly. Gathering the music and sound effect is also a hassle because you'll need to find these assets with license that can be used commercially (I assume). Simply said, I don't want to get in trouble. There's literally not much I can do about this, because it's an essential process to go through each dialogue and add sounds to it one by one... Unless! Unless I add the sound effects and music as I write the dialogues. I thought about it before, and never once act on it. Maybe I assume that doing so will disrupt my workflow, and very likely might. I'll try it out in the chapter 3 and see how it goes.

- Coding

Coding usually isn't a big concern. A quick google can show me pretty much how to do anything. I really only have to dig deep if I want to do something like a phone system or mini games... I'm still a big newbie for all this coding stuff but I'm willing to learn if it means I can make a gambling mini-game. Anyway, still have to do the gallery and phone system. Otherwise, not much to talk about coding. I'm not planning to do any crazy shit with this game. Just want to keep everything simple, and cinema.

Development sneak peak

One member of us suggested that I should share my insight on how action scenes are crafted for Sort of Justice. I believe it's an excellent idea and a great opportunity to further connect with my members. I'll release another post about this process later in the future when I'm creating one of the action scene in chapter 3, which I can use as an example to demonstrate the techniques and methods I used. I'll also incorporate images to visualise various points to make it easier to understand. Hurray, another yap session!

Steam release

I should clarify for any steam players, that, Sort of Justice will not release update on every chapter to Steam. In the future, steam version will be released as "seasons", comprised of multiple chapters, likely 4-5. This will be sold as DLC for the existing steam page.

Acknowledgement

Thank you for all of my supporters and fellow developers for the continual support. Whether you are members who had been here since day one or the newly members who have joined only recently, I hold dear to all of you. Special shout-out to my patrons who decided to donate their hard-earned money just to see this game thrive. I'll try my best to make sure that this game's ending sees the light of day. That's it for this yap session, I'll see you in the next one!