How do you guys orient yourself to the medium to reduce the amount of renders needed for a scene?

[Lugosi]

Maybe a simple question but I wanted to ask in case any of you were consciously thinking about these things. As I am finally chugging out renders I noticed that I can't really afford to render everything I originally wanted for the sake of time. When first writing the script I had it in my mind almost like a movie (ie. A contentious stream of visual information).

Let me give an example of what I am talking about in case I am unclear. The scene -> MC is hurt, Li brings him to the couch.

What I had envisioned:

MC follows LI into living room -> MC walks to couch -> LI walks over to other chair to grab a pillow -> MC sits down -> LI grabs pillow -> MC takes off shoe -> LI is now closer bringing the pillow -> MC takes off other shoe -> LI now is close with pillow -> MC starts to lay down -> LI puts pillow under his head -> MC finally lays down.

This strategy is almost completely unsustainable. If you employ a bit of creative and intelligent planning, you could likely accomplish that sequence is 4-5 renders. The original chain of events is planned as if you were making a film and frankly that sequence is way more realistic, buts that's 12+ renders of lightning and positioning without any words. My script already has like 9k words. If I keep this up I'll never release the damn thing.

This is rendering in DAZ btw but this would apply to any render program that requires more involved scene setup. I don't imagine people who use say Honey Select would have this problem.

 

[MidnightArrow]

I used Blender for v0.1 of my game, so it isn't unsustainable at all if you use the right program. Blender works great for these kinds of involved "cinematic" scenes, but the setup required is too involved for the "apply presets and iterate quickly" Daz model that works well for Ren'py dialogue scenes. So I plan to mix both programs next time. Obviously I'm still going to do all my posing in Blender, because Daz posing is dogshit.

If you do want to cut the number of renders down, then you need to keep in mind this is a visual novel. Emphasize dialogue/monologue over visuals, and save the event stills for the really big one, like the love interest putting the pillow under his head. Does it matter if we see him take his shoes off?

 

[anne O'nymous]

MC follows LI into living room -> MC walks to couch -> LI walks over to other chair to grab a pillow -> MC sits down -> LI grabs pillow -> MC takes off shoe -> LI is now closer bringing the pillow -> MC takes off other shoe -> LI now is close with pillow -> MC starts to lay down -> LI puts pillow under his head -> MC finally lays down.

Well, You achieve that by mixing relevant renders with relevant dialog lines.

Having the whole "animated" scene in your head is good, it help you pace the dialogs and actions. But it's just a help, and in the end you should only keep the renders that are effectively significant :

• We are not idiots, we know that people don't teleport, so no need to show MC walking to the couch or sitting down.
• We don't need to know where the LI found the pillow, we know she haven't use magic to invoke it.
• We also know that he will not remove just one shoes, so no reason to show him removing one, then the other.
• We still not idiots, we know that the pillow will not magically appear behind MC's back, so no need to show LI putting it.

In the end, it left five renders, that you punctuate with dialogs to mark the time passing:
MC & LI are in the living room
LI "sit down, it's not good for you to stay up for so long."
LI pick up a pillow
MC "Yes mom..."
MC is on the couch taking off his shoes
LI "Take this, instead of trying to be a smart ass."
LI give the pillow to MC
MC "Thanks."
MC is laying down on the couch.

 

[8InchFloppyDick]

Well, You achieve that by mixing relevant renders with relevant dialog lines.

And all of the rest of Anne's response is spot on. Chapeau!

 

[Lugosi]

Well, You achieve that by mixing relevant renders with relevant dialog lines.

Having the whole "animated" scene in your head is good, it help you pace the dialogs and actions. But it's just a help, and in the end you should only keep the renders that are effectively significant :

• We are not idiots, we know that people don't teleport, so no need to show MC walking to the couch or sitting down.
• We don't need to know where the LI found the pillow, we know she haven't use magic to invoke it.
• We also know that he will not remove just one shoes, so no reason to show him removing one, then the other.
• We still not idiots, we know that the pillow will not magically appear behind MC's back, so no need to show LI putting it.

In the end, it left five renders, that you punctuate with dialogs to mark the time passing:
MC & LI are in the living room
LI "sit down, it's not good for you to stay up for so long."
LI pick up a pillow
MC "Yes mom..."
MC is on the couch taking off his shoes
LI "Take this, instead of trying to be a smart ass."
LI give the pillow to MC
MC "Thanks."
MC is laying down on the couch.

Thank you! Logically I came to the conclusion that I had to figure out the best way of appropriately "filling in the gaps" but having it laid out like this is very helpful. Appreciate it.

 

[TessaXYZ]

I have a few rules I'm loosely following.

1) Does the dialogue allow the reader to infer a behavior? If not, add a render that shows the behavior.
2) Does a behavior or action need to be presented in semi-real-time pacing to have the right effect for the narrative? If so, render additional images as needed. Example: I have a scene where a character is being followed. This is new to the reader, and I have a scripted sequence of 4 renders where a car rolls up slowly from off-camera.
3) Maximum of four lines of dialogue per render.
4) Does a line of dialogue require immediate expressiveness? If so, show in a new render. For example, the character learns something shocking. Show that shocked face immediately.
5) For time jumps, would there be any conversation or scene in that time that definitely requires exposition instead of inference? If so, either develop that scene or add some "train of thought" contemplation about it.
6) Plan out the entire scene and all its dialogue and actions before starting to render. Don't just render without a plan. I have a google sheet written out in script form. E.g.:

Scene 3. Image 1: Wide shot of both characters entering the house. The lights are on, but no one is in the foyer. Second character is calling out to let everyone know they've arrived. Alexis: "We're here, ladies!"

 

[moose_puck]

Since I still can't afford a decent GPU, I've been forced to find all sorts of time-saving shortcuts with my project.

Here's a few things I do that save a lot of time running renders...

1. Set the scene up completely before you start to render. That includes lights, cameras, all props etc. Because a lot of what i do to save time involves spot renders and composting images, I need to make sure the baseline is perfect, so that my renders don't have a different look. About the only thing I change between renders is the characters themselves and usually 3 or 4 spot lights. I have the spots set to how I like them and I just move them around for each different pose. I used fixed ghost lights for most of the interior environment shots.

2. Once you have your scene composed, save that DUF and keep it unchanged. Then create new dufs for all the changes as you go and pose your characters. You know there's going to be stuff you screwed up and don't see until you review the completed render, so make sure you save every duf for even minor changes, so you can then just do a spot render to fix those items.

3. I've learned that environments take a huge amount of time to render, while the genesis characters are actually pretty fast. So for example, I have a scene with 2 gen 8 characters in a bedroom, with the usual background stuff. If I run that render at 1080p all in one go, I can easily take 3 hours and 3000 or so iterations to get it good enough quality. but if I do the background first, that only takes an hour, then the characters only take 20-30 minutes for 3000+ iterations. I then composite those images together in post production. I've not only saved myself 1.5 hours of render time, but I can play with the lighting on the characters without changing the environment at all. I particularly like to use a spot behind the characters to get that nice highlight on the curved outside of the characters and the highlights in the hair. You can't do that if you render the scene all at once.

4. Also, by setting your cameras and background up in advance, you now only need to render the characters and that can be very fast. Occasionally you might need to make a shadow or something, and even then, you can just run a spot render over the area you need to change, which is much faster then doing the whole frame again.

5. I mention spot renders constantly... that's because it's probably the best tool in DAZ for speeding your production up. Make a mistake and have a finger poking into someones side? No need to do the whole render over. Just move the finger and spot render that area. Could be as fast as only a few seconds, compared to 30 mins+ to run the whole frame again. Also, use spot renders and small movements and changes in character poses to give them life - to many devs just paste a mouth open or close to show someone talking. That looks totally unrealistic IMO. What I do is I try and make a small head movement and maybe have a hand gesture a little between each dialogue. It only takes a minute to do those spot renders and I find it looks so much more real then the wooden mouth opening and closing.

6. Finally to add to what was mentioned above about your workflow... I agree that it's unsustainable to make a render for every minor action. Instead, i would render the most "expressive" scenes and then use dialogue to bridge the transitions. When using dialogue though, don't describe what is on the screen visually... after all, a picture is worth a thousand words, right? Instead, use the dialogue to lead into or describe something that may have occurred between scenes.


Render 1 - MC and LI entering living room together
dialogue - LI "Why don't you lie down on the couch and I'll go grab you a pillow."
Render 2 - MC sitting down in process of removing shoes (LI offscreen)
dialogue - MC "Thanks. My neck is killing me."
Render 3 - LI putting pillow under MC's head as he starts to lie back
dialogue - LI "Here you go, just lie back and relax."
Render 4 (optional) - LI sitting beside MC on couch, maybe doing a gentle caress of his forehead or something...

The 3 renders above pretty much cover your whole scene. Render 4 is an optional one that I would do, depending on the context of your scene, which we don't know of course. I'd do it to set the mood... maybe highlight the feelings the LI has for the MC in the concern on their face as they watch the MC lying down.

One final thought, on how I am approaching the process for writing my VN... I don't write my dialogue until the very end of my workflow, when I am doing the coding in Ren'py. Oh, I'll sometimes put a note in my excel sheet when i do the rendering. But that's just a "post it" style note for something I don't want to forget, like a nice joke or witty remark I want to use. Instead, what I do is I work off a screenplay of sorts. In fact, I write it based on what I have seen for TV screenplays. where each scene is just described in point form. Example "Scene 8 - Coffee Shop - MC and LI go get coffee. MC flirts with waitress. LI gets pissed. etc.." I make the renders up based on what I think will happen and I make sure to add some basic dialogue back and forth renders.

When all my renders are done, I then sit down and start coding each scene and writing the dialogue as I go. This is by far, the most enjoyable part of the process for me! I usually put some music on, have my Atom editor up on one monitor, my project folders on the other monitor and I preview the scenes on my TV. Then I put my writers hat on and have a blast. Finally, I let the scene sit for a few days then go back and run it again. Sometimes you notice things that don't work after you have taken a break from it for a while. I learned that from my days doing art in highschool.

 

[TessaXYZ]

One final thought, on how I am approaching the process for writing my VN... I don't write my dialogue until the very end of my workflow, when I am doing the coding in Ren'py. Oh, I'll sometimes put a note in my excel sheet when i do the rendering. But that's just a "post it" style note for something I don't want to forget, like a nice joke or witty remark I want to use. Instead, what I do is I work off a screenplay of sorts. In fact, I write it based on what I have seen for TV screenplays. where each scene is just described in point form. Example "Scene 8 - Coffee Shop - MC and LI go get coffee. MC flirts with waitress. LI gets pissed. etc.." I make the renders up based on what I think will happen and I make sure to add some basic dialogue back and forth renders.

When all my renders are done, I then sit down and start coding each scene and writing the dialogue as I go. This is by far, the most enjoyable part of the process for me! I usually put some music on, have my Atom editor up on one monitor, my project folders on the other monitor and I preview the scenes on my TV. Then I put my writers hat on and have a blast. Finally, I let the scene sit for a few days then go back and run it again. Sometimes you notice things that don't work after you have taken a break from it for a while. I learned that from my days doing art in highschool.

That's what I did initially, but found that the dialogue I came up with often felt better with expressions or poses that weren't originally set in the renders, or that it necessitated additional renders to feel right. So now I have all the dialogue written out before the renders, which actually kinda makes setting up scenes more straightforward for me by knowing exactly how that render is being used. Not saying your method is wrong, just interesting how different people plan and execute their story.

And thanks for the heads up about spot rendering, wasn't aware that's a thing -- will have to look into it.

 

[moose_puck]

That's what I did initially, but found that the dialogue I came up with often felt better with expressions or poses that weren't originally set in the renders, or that it necessitated additional renders to feel right. So now I have all the dialogue written out before the renders, which actually kinda makes setting up scenes more straightforward for me by knowing exactly how that render is being used. Not saying your method is wrong, just interesting how different people plan and execute their story.

And thanks for the heads up about spot rendering, wasn't aware that's a thing -- will have to look into it.

Well, I do have to sometimes go back and do more renders, but that's why it's good to have every DUF saved. Then it's easy to just compose and render the new scene. And because coding and working with Atom & Excel isn't very memory intensive, I can continue to code and write dialogue while those renders are running. For me, I just find the writing part is easier and more fluid when I am doing it while looking at the finished renders. In fact, I'll often have the folder with all the renders for the current scene I am working on, up on my TV screen in extra-large icon size. I'll sit back and look at the images as I am writing, with some music streaming in the background.

I guess everyone has to find what works for them. I'm thinking about trying to do one of my next scenes after I take some kick ass balanced edibles I have been using lately. They are 10mg of THC and 10mg of CBD and man! - do they make me mellow out! Which is a good thing, because I am pretty hyper alert most of the time and sometimes have trouble focusing on the writing without getting distracted all the time.

 

[JustXThings]

I use black scenes often for transitions. For instance, something similar to what you described.

Render 1) Guy hurt and LI putting his arm around her neck, waist, whatever

LI -> "Wow... you look like you have seen better days."

LI -> "Let me help you to the couch."

Render 2) black screen

LI -> "Lay down..."

LI -> "Wait a moment... I'll grab an extra pillow."

Render 3) Render of the guy on the couch, pillow set and LI with her hand in the pillow fixing it

LI -> "Better ?"

But I also do tons of in-between renders to make scenes livelier. I have several renders of a girl cooking, just moving the frying pan around to make her feel alive. Gosh... I even have scenes of 3/4 renders without dialog.

Talking about workflows, mine is pretty simple.

1) Describe the scenario in a notepad
(2a, 2b can be done in any order)
2a) Write a draft of the dialog (including branches with "pass")
2b) Render key frames (usually 3, 4 per scene. all other renders will have similar perspective and lighting to keep visual cohesion)
3) Once I have the full scene with the 3/4 key frames. I start polishing the dialog and matching the in-between renders accordingly.

Pretty much that.

Final note. It is worth expending time playing with the lighting, removing walls, etc. to reduce the render times of your key frames. As well as removing any non visible parts of the environment or models (clothing, hair, browns, etc.). Any in-betweens will benefit from that. But careful there or you will end up with bald shadows.

 

[shabadu]

I'm going to make an assumption here that you're doing unique renders for every scene in your game. If so, then maybe you could try implementing a sprite-based system. This technique is way more common with 2DCG stuff, but that doesn't mean you can't use it yourself.

In case you don't know what I'm talking about, in a sprite-based system you would split up your scenes between a static background render and layer on transparent cutouts of characters done in something like a 3/4 bust format. You would also have different character sprites to cover different emotions, such as happy or embarrassed, and swap them out for what the dialog / story calls for. You can also go further and have sprites for different outfits, maybe change the lighting for time of day, and so on.

The major benefit here would be that, once you have this set up, a lot of the dialog-only scenes are now taken care of and don't need any specific renders themselves. Sure, you would need to go through the process of getting it all working, but that initial effort will pay off the more you work on your game and the bigger it becomes. It would also free up more of your time and GPU for doing full renders of the big important scenes. I'm going to make another assumption that you're using Renpy, so I'll also note that there is built-in functionality for this exact thing so getting it to work on that side should be relatively hassle-free.

Now, this isn't some perfect fix for your situation. It may well be the case that this just doesn't fit with what you want to accomplish, but it is a potential way of approaching your game that can help you to better manage your resources. It would also be a good litmus test of sorts when you approach how to go about some of the scenes and story moments - you can try to convey what's going on through just the sprites and dialog, and then only expand out to full scene renders when needed.

As a quick example, in incest-themed games there always seems to be at least one scene of the family having a meal together. It's almost always visually boring, with characters sitting at the table and iterating through relatively similar poses until the scene ends after 10+ images that are 90% the same as each other. This is, in my opinion, a perfect time to use sprites. In the abandoned game Love and Submission, there is just such a scene. Let me be clear, I'm not saying that the game is bad, or that the developer veqvil didn't do it the right way or anything like that, just pulling an example that's fresh in my mind.



Don't mind the compression and quality, that's on me, but there are 17 images I combined in that gif. That's 17 unique renders that could have been skipped if veqvil had used sprites. In fact, this whole scene has total of 30 images associated with it, but I cut out the ones showing the MC walking to the table at the beginning and where the mom character stands up and leaves at the end. Again, let me just say that there is nothing wrong with doing it this way, but to me that is a lot of time and effort spent where it wasn't needed.

 

[coffeeaddicted]

I'm going to make an assumption here that you're doing unique renders for every scene in your game. If so, then maybe you could try implementing a sprite-based system. This technique is way more common with 2DCG stuff, but that doesn't mean you can't use it yourself.

In case you don't know what I'm talking about, in a sprite-based system you would split up your scenes between a static background render and layer on transparent cutouts of characters done in something like a 3/4 bust format. You would also have different character sprites to cover different emotions, such as happy or embarrassed, and swap them out for what the dialog / story calls for. You can also go further and have sprites for different outfits, maybe change the lighting for time of day, and so on.

The major benefit here would be that, once you have this set up, a lot of the dialog-only scenes are now taken care of and don't need any specific renders themselves. Sure, you would need to go through the process of getting it all working, but that initial effort will pay off the more you work on your game and the bigger it becomes. It would also free up more of your time and GPU for doing full renders of the big important scenes. I'm going to make another assumption that you're using Renpy, so I'll also note that there is built-in functionality for this exact thing so getting it to work on that side should be relatively hassle-free.

Now, this isn't some perfect fix for your situation. It may well be the case that this just doesn't fit with what you want to accomplish, but it is a potential way of approaching your game that can help you to better manage your resources. It would also be a good litmus test of sorts when you approach how to go about some of the scenes and story moments - you can try to convey what's going on through just the sprites and dialog, and then only expand out to full scene renders when needed.

As a quick example, in incest-themed games there always seems to be at least one scene of the family having a meal together. It's almost always visually boring, with characters sitting at the table and iterating through relatively similar poses until the scene ends after 10+ images that are 90% the same as each other. This is, in my opinion, a perfect time to use sprites. In the abandoned game Love and Submission, there is just such a scene. Let me be clear, I'm not saying that the game is bad, or that the developer veqvil didn't do it the right way or anything like that, just pulling an example that's fresh in my mind.

View attachment 2083796

Don't mind the compression and quality, that's on me, but there are 17 images I combined in that gif. That's 17 unique renders that could have been skipped if veqvil had used sprites. In fact, this whole scene has total of 30 images associated with it, but I cut out the ones showing the MC walking to the table at the beginning and where the mom character stands up and leaves at the end. Again, let me just say that there is nothing wrong with doing it this way, but to me that is a lot of time and effort spent where it wasn't needed.

Agree.

I think sprites are so great.
One game that used it well (imho) is Light of my Life.
There are surely others.

All you need to do is to setup the light in the scene and use the lights for the sprites alone. The renders will be quick which helps.
The problem is not to over do it. To find the right balance.

 

[coffeeaddicted]

Maybe a simple question but I wanted to ask in case any of you were consciously thinking about these things. As I am finally chugging out renders I noticed that I can't really afford to render everything I originally wanted for the sake of time. When first writing the script I had it in my mind almost like a movie (ie. A contentious stream of visual information).

Let me give an example of what I am talking about in case I am unclear. The scene -> MC is hurt, Li brings him to the couch.

What I had envisioned:

MC follows LI into living room -> MC walks to couch -> LI walks over to other chair to grab a pillow -> MC sits down -> LI grabs pillow -> MC takes off shoe -> LI is now closer bringing the pillow -> MC takes off other shoe -> LI now is close with pillow -> MC starts to lay down -> LI puts pillow under his head -> MC finally lays down.

This strategy is almost completely unsustainable. If you employ a bit of creative and intelligent planning, you could likely accomplish that sequence is 4-5 renders. The original chain of events is planned as if you were making a film and frankly that sequence is way more realistic, buts that's 12+ renders of lightning and positioning without any words. My script already has like 9k words. If I keep this up I'll never release the damn thing.

This is rendering in DAZ btw but this would apply to any render program that requires more involved scene setup. I don't imagine people who use say Honey Select would have this problem.

I think it boils down to, do i really need to show everything.
Some things can be left to the imagination other need to be shown.

Like, is the pillow essential to be grabbed or not?

I made a scene where MC checks the rooms for the arrivals and turns around to switch off the lights to the rooms.
There are two rooms.
So i end up with 5 renders.

But the most important lesson i took for me is, even though i render something, i decide late on that this slide isn't worth it and take it out. Keep the scene in DAZ, you may need it later again.
But for your story it may not be needed.
I think i have to ask myself always if it is essential to have this particular scene. What do i want to show?

In the beginning i ended up with a lot of slides but it was more like a movie. So i had to make changes and take scenes out because they weren't needed to tell the story.

Just my take.