VN Ren'Py Sucker Punch (AI) (Story Driven) [Development Thread]

[1nsomniac22]

Greetings F95zoners. As a recent emigree from itch.io (prior to the Adult Content apocalypse), this thread will be the Devlog for my new project: Sucker Punch. I am very pleased (and as always, a little nervous) to release a new effort to the world at large. I invite you to take a look and see if this story interests you.

TL;DR, new game, Ren'Py, AI art. First chapter is available here:

You must be registered to see the links

.​

Sucker Punch is the Adult Visual Novel I have wanted to create for a while. Where my first work (Workin' Out) was a Kinetic Novel with no user interaction, this is my best effort to write a compelling narrative where the players choices dictate how the story progresses and relationships develop (or don't).

Fighting to tell a story
​

This project started with the concept of having an action component to the story - the fight mini-game. It's (admittedly) a very simple game mechanic, but I hope it adds something to the user experience more than just clicking through dialogues, selecting menu items and looking at pictures (not that there's anything wrong with that). I wanted to create an experience that involves the player and generates tension, hopefully the first fight mechanic provides that. Remember, fights aren't always a win or die scenario, you don't have to win every fight to progress through the story (only Critical Fights). There are some dialogues/choices that only happen when fights are lost.

World building
​

Sucker Punch is set in a gritty current day world. You live in the slums of a nameless city working a dead end job. You're poor and desperate to get out of the current situation. The story starts on the eve of your final amateur fight; win this fight and go professional. But something terrible happens and you find yourself scrambling to pick up the pieces.

The characters you meet in this world are also struggling to survive. I am trying to develop characters with their own stories, motivations, and narratives. This means that each character you meet is usually broken in some way - some more obviously than other. But, like the fight system, I'm trying to create a storyline with multiple routes that aren't just- yes: have sex/ no: skip the sex scene or do you have enough love points to see this scene? I admit, this is difficult- the technical challenge is manageable in Ren'Py (I work professionally with Python, so I've got a good enough understanding of how to code this), but developing these storylines so they offer distinct branches, but still return to the main story arc seamlessly is challenging - in a good way.

This Adult Visual Novel is told through images generated by Stable Diffusion (not everyone's cup o' tea). I've put a lot of effort in developing a style that isn't the default Pony (or other basemodel of your choice). I've found those basemodels generate images that are too smooth (for lack of a better term) or conversely, hyper-realistic images which can be stunning, but I think that dip too far into Uncanny Valley. This is an artistic/stylistic choice, I'm interested in seeing how far I can go using local AI generated imagery. I'm aware there are critics of AI generated content... I'm one of them. I am not a fan of the churned out basemodel slop that's flooding every forum and medium. There are a lot of guilty parties publishing stuff that I feel gives the (AI gen art) technology a bad wrap. However, AI generated art is a new medium that I'm more interested in working in than Blender, Poser, or Daz.

The cast in the Prologue and Round 1 you will interact with in Sucker Punch​

Ashleigh is a nurse who cares for the player after the fight.

Spoiler: A chance encounter reintroduces her...

You don't have permission to view the spoiler content. Log in or register now.

Rikton is the owner of the Dollhouse club.

Spoiler: You become a bouncer at the club...

You don't have permission to view the spoiler content. Log in or register now.

Lilly is a naive girl who works in the club.

Spoiler: She has a care-free personality...

You don't have permission to view the spoiler content. Log in or register now.

Elizabeth is an exotic dancer in the club.

Spoiler: She has an unnatural intuition...

You don't have permission to view the spoiler content. Log in or register now.

If you've read this far - thank you! I am completing the final editing of Round 1 and will be publishing it shortly. Obviously this VN is intended for adults only. It includes scenes of:

• Smoking
• Alcohol use/Alcoholism
• Illegal drug use
• Sexual situations including coerced sex, prostitution, BDSM, male and female dominance
• Emotional and physical trauma, mental illness, depression and suicide

They are not always optional, view at your own risk and in compliance with your local laws. You have been informed.

Questions/Answers:

1. Are the fighting mini games skippable? Yes, in the game Options menu you will be able to suspend/resume them at any time.
2. Is there a harem route? No. You can pursue multiple LI's, but they're not going to be one happy octo-thruple.
3. Are there gay scenes? Not planned, the MC is heterosexual male. In the femdom route there will be exploration of anal play, pegging and other similar topics, but no siffication - that doesn't really fit the MC character.
4. Are there lesbian scenes? I do have a story route planned that involves a lesbian relationship. It will be optional to pursue that route.
5. Will there be scenes of LI's with other men? Yes, but not as a voluntary relationship. Think more coerced sex and exploitation.
6. Will there be NTR content? No, not by the common definition. See answer 4 for more context.

 

[Winterfire]

Will you clean your art? Almost all of the faces are broken, and they all seem to be using a different style... Especially the backgrounds, in comparison to the characters. I'd focus on a single style for consistency.

 

[duckydoodoo]

im more concerned with are mini games skippable, is there free roam or sandbox?

as for the dark elements, its all in the story telling. for the most part ill try any type of game content and see how its done then decide if its for me. but i tend to avoid sandbox, i dont care much for free roam but can stomach some, and i loathe mini games and if i cant skip them i skip the game. 9/10 times i cant complete the mini game. im disabled and my reaction time is shit. had to give up playing video games years ago because of it.

 

[1nsomniac22]

Will you clean your art? Almost all of the faces are broken, and they all seem to be using a different style... Especially the backgrounds, in comparison to the characters. I'd focus on a single style for consistency.

The backgrounds and characters use different base models due to the limitation of the Pony model and it's extremely annoying lack of understanding of non-sexual non-character tokens. Using different models for the characters and backgrounds is something I've done for my other project as well.

Could you give a bit more information about you are seeing that is broken in the faces? I'd like to know if there is a flaw in the LoRA's I've trained.

im more concerned with are mini games skippable, is there free roam or sandbox?

Sorry, I haven't planned to implement free roaming/sandbox due to my own limitations of writing. I'm afraid it would be too difficult to keep the storylines advancing with a truly open experience. Early in the development cycle I was considering having a more open story using the day/action points model, but abandoned that avenue of development.

I'm sorry to say the minigames are not skippable as I've implemented them; I'm afraid this VN would be quite frustrating for you in it's current form. I will look into how difficult it would be to add an option to prevent the minigame/substituting a fight won/lost selection menu. Technically it should be easy, but I don't think I would have it in place before my planned ship date.

Thank you for the suggestion, I don't want to put up unnecessary barriers to the VN and I hadn't considered that some players wouldn't actually be able to complete the minigame.

 

[Winterfire]

The backgrounds and characters use different base models due to the limitation of the Pony model and it's extremely annoying lack of understanding of non-sexual non-character tokens. Using different models for the characters and backgrounds is something I've done for my other project as well.

Could you give a bit more information about you are seeing that is broken in the faces? I'd like to know if there is a flaw in the LoRA's I've trained.

 

[Maviarab]

Certainly sounds interesting...mini-games..uugghh..why do people keep investing time coding this shit when they could spend more time actually devloping their story, more writing and rendering...bnut no...gotta code in these games. It's a VN...not a 'game'....at least your dropped the sandbox idea...something at least heh.

Depending on how it's actually implemented I could grim and bare it, if the story is compelling enough. Look forward to seeing where/how it goes.

Good luck Dev!

(Might want to also addres in this threead what the usual questions will be...possible harem?...lesbianism? (skippable)...see LI's in compromising situations with other men (NTR)...blah blah blah...you know what's coming when it releases heh).

 

[1nsomniac22]

im more concerned with are mini games skippable, is there free roam or sandbox?

I took a second look at the code and think I have a solution for skipping the fight minigames. There will be a selection in the Options menu: Skip Fight Games. If true instead of the minigame you'll be presented a Win/Lose menu. I'm doing more testing and need to add some final formatting, but this feature will make the release date.

Again, thanks for the suggestion!

 

[1nsomniac22]

(Might want to also addres in this threead what the usual questions will be...possible harem?...lesbianism? (skippable)...see LI's in compromising situations with other men (NTR)...blah blah blah...you know what's coming when it releases heh).

Thanks for the heads up, will update OP with those usual questions.

 

[1nsomniac22]

TL;DR, I plan to ship the Prologue & Chapter 1 in a few days.

I have completed edits of the Prologue and Chapter 1 manuscript. I got stuck in a feature creep/bug chasing doom loop over the weekend, but I'm pretty sure I have that resolved now. All that remains is a final clean up of the images and a final bug check/review.

I had planned on wrapping the chapter where it was, but last week I went back through a few scenes and added additional renders/SFX's - and I am happy that I did; I feel it really adds to one aspect of the story... at the cost of two days I should have been editing.

At the same time, I'm also redoing a bunch of the character renders. Most of the prologue and many of the chapter 1 renders were originally done with the Automatic1111 UI. Now that I'm using Invoke AI and its better inpainting, I feel like the final product is much better. Also, it doesn't help that the Lilly LoRA was redone as I was completing chapter 1; so there is a significant improvement in character looks and fidelity when I redo them.

Original:

Reworked:

All future chapters will be generated exclusively with Invoke AI, so hopefully the reworks will be minimal from here out. This might sound like a cop-out, but while the visuals are very important in a VN (derp), using SD to generate them is more art than science. There's always going to be that little bit of inconsistency or something not quite right. Maybe the newer gen tools are doing a better job of this -looking at you Nano Banana - but apparently it's completely NSFW gimped by Google/Alphabet, never mind my requirement that the model runs locally on my hardware... but that's a topic for another rant at another time.

Regardless, since time is finite, the time I focus on the visuals is time I can't spend doing another editing pass to try to keep the dialogue/narration tight - which is the other big part of a Visual Novel. I'm not complaining, nor making excuses. I'm writing this post to keep myself accountable.

Obviously, I want to produce the best quality VN that I can, but I've given the project the requirement (limitation) of using AI/SD generated visuals, at some point I need to accept what is done and stop obsessing about those last 5% issues. This is the challenge in deciding when something is ready to ship.

And... I really need to stop adding features... most recently I got side-racked into a feature to read the soundtrack info from a JSON file and then reformat it into the About text. Yes, in the long run it will make it much easier for me to document and attribute the songs I've included in the VN, but... not right before shipping -_-.

Shipping will be on Windows/Linux, Mac and Android platforms. I was pleased in local testing that the mini game worked on the phone, and actually played better in that format due to the touchscreen, it's more intuitive than using the mouse. Someday I may test on a touch screen...

This release will consist of:

• Prologue [Day 0 - Day 15]
• Round 1 [Day 15 - Day 29]

I have my fingers crossed that I can wrap up the last few re-renders in another day, then crank through the last few test runs. If you've read all of this and are waiting on the release, thank you! If anything changes on the release schedule I'll post updates, but in the meanwhile I continue to close out this last sprint.

Regards,
-Insomniac

 

[eevkyi]

I really like the setting as a whole, but Ashleigh and Lilly look too young. Unless you change their designs, it’s almost certain your game will be banned on non-japanese platforms and websites.

 

[1nsomniac22]

I really like the setting as a whole, but Ashleigh and Lilly look too young. Unless you change their designs, it’s almost certain your game will be banned on non-japanese platforms and websites.

Thank you for comment and I appreciate the heads-up!

I have been doing a bit of clean-up and fine tuning on the final renders to make sure I don't run afoul of any of the rules. Never have I supplied the AI a prompt or used a LoRA for rendering the female characters of a(ny) specific age, but that's immaterial to the final output, which is what the game will be judged on. I'm afraid, there is a bit of ambiguity when defining fully synthetic actors - the Gwendolyn model from Workin' Out went through a similar development process, and shows similar age(less) characteristics.

FWIW, I have already published both Workin' Out and the prologue chapter of Sucker Punch on itch.io last and earlier this year - before the apocalypse (but have since removed S.P. during the rework/preparation for this release). Unfortunately, we're in the post-apocalypse era, so yea, unfortunately, there's definitely a possibility.

The game disclaimer states that:

All characters portrayed are fictional and at least 18 years old.

The character bibliographies (for what little it's worth) state that:

• Ashleigh is 19 years old
• Lilly is 24 years old
• Elizabeth is 26 years old

There will never be scenes where the female characters are in any kind of school setting, so there should be no contextual inference made about the character ages (I refuse to play the High School that is a University semantic game).

I'll do the best I can to release the VN as true to the original concept/manuscript as I can. I have designed the characters with their specific looks for reasons based on their particular history. If if I have to adjust things along the way I certainly will.

 

[1nsomniac22]

You must be registered to see the links

TL;DR - game page is live, game is available for download.

The last sprint is complete, but it was a bit of a fight to close out the last tasks.

1. I introduced a major bug with the changes to the music attribution system that was only visible when the final packaged archives were created/accessed. That might be an interesting post in it's own right.
2. After fixing that bug I needed to complete another two test runs to give myself peace of mind.
3. Completed another editing pass on the manuscript; I hope I've caught all the typo-s & grammatical errors.
4. Got wrapped up in another review of renders and fixing things that were wrong.

Now that chapter 1 is published, I feel comfortable sharing the story flowcharts for the prologue and chapter 1:

Prologue:


Chapter 1:


The increase in complexity from the Prologue to Chapter 1 was significant - which definitely made the development cycle a lot harder than I expected. I've got some ideas how to deal with this for future chapters (which will only increase complexity and branching), so I hope I can manage a quarterly release schedule for new chapters.

Regardless, those are future concerns. Right now I'm taking a breath and keeping my eye open for reports of any missed bugs/launch issues.

If you're so inclined, please give the game a try. I appreciate any critique or suggestions.

Cheers,
-Insomniac

note: de-obfuscated the flowcharts; was kinda spoilerish.

 

[1nsomniac22]

Post Launch Update

TL;DR - the weeks after launch were busy dealing with issues and other things, work on Chapter 2 is ongoing.

Apologies of a sort, I want to write at least 1 update per week to the devlog, but the run-up to the launch and then post launch things kept me busy. I found a few small issues that required immediate fixing, and one that was kinda stupid - I shortchanged the player $500 for the last two nights of work at the club. OOPS... I mean money shouldn't be that tight by the end of chapter 1, but still - it's the principal that counts.

I am fully into the writing and development of Chapter 2 of Sucker Punch, here is the story flowchart as I have it currently (but obviously it may change further):

Chapter 2:


This chapter will see the introduction of the final two "main" LI's, which will set the stage for the remainder of the story. I'll detail them more in later devlogs, but these introductions will be of an even darker theme than the earlier chapter, I feel going that route is necessary to tell the story as I've envisioned it.

I am working out the best mechanic to force the player to chose between Leigh and Ash for their next encounter - I want to give the player more control how that relationship develops (Leigh & Ash have kind of thrown themselves at the player in chapter 1), and I want there to be consequence for the decision. I just need to figure out a clever way to present the choice to the player.

Also, the second LI relationship begins to develop in chapter 2 - if the player's chosen to pursue it. I'm looking forward to more fully developing Lilly's character, background and motivations. I have a little bit of programming fun to explore in one of the scenes I've planned, so that will either be amusing or frustrating - depending on how stupid I am that day.

On a technical note - I've been exploring a different workflow for generating the images through Stable Diffusion.

Typically there are three main operations and 1 cleanup that need to occur for most of the single subject images. Multisubject images need additional image layout work (frequently outside SD using image manipulation tools to compose and align the various subjects) :

1. Image layout
2. Face/expression refinement
3. Body & clothes refinement
4. final cleanup

The initial image layout can be sometimes very challenging if I am trying to describe a concept to SD that the model doesn't understand. In those cases I've resorted to using public images to define control net model pose skeletons, with limited success (see the judo throw and front kick in technical training).

Once the image is blocked out I use a second, highly focused pass on the face(s) using inpainting and the appropriate LoRA, which usually works - but as with all AI generated images, there's an unfortunate amount of drift. AI tools of the SD 1.5 generation always seem to have issues with character consistency (as well as hands).

Recently, I've been using the control net lineart model to revisit some of the images that shipped in chapter 1 that I wasn't happy with.

For example, the Ashleigh tank top images faces seem a bit too different given the the LoRA strength and expression I was trying to render. But, using those images as a base, processing them with the lineart model solves the Image Layout and blocking challenge and lets me generate an entirely fresh image using the Ashleigh LoRA with appropriate expression prompts.

Image shipped in Chapter 1:

It's not a bad image, but I don't think the face is consistent enough with Ash from earlier in the chapter, and the expression isn't 100% what I was going for. And it seems a little low detail overall.

Processing that image with the lineart filter gives this:


And then creating a whole new image using that control base with the LoRA and the remaining steps (blocking, face & body detail) give this (note, not final cleaned image):

And this I feel is much truer to the other Ashleigh images in the story, while preserving the pose and blocking of the original. And it is a (relatively) quick job to process through.

And the lineart model is working for both de-novo images as well as reprocesses. For example, here's something I have planned for Chapter 2 where I took a few quick renders of the concept, lineart-ed it and then did the detailer steps (note, this isn't final either, it hasn't been detail checked or cleaned, it's more of a storyboard for me):

But it was a much faster process to go from initial image concept to this stage than my prior workflow.

All that said, I'm going to spend a bit of this development cycle going back through the earlier images and making sure they're consistent. I'll try not to let the reworks impact the overall development timeline - where I plan to ship Chapter 2 sometime around the end of the year.

All that said, hope this update makes up for my earlier deficiency.
Regards,
-Insomniac

 

[1nsomniac22]

What happens when a side character becomes more?

TL;DR - adding another main LI, reconsidering the development scheduled.

In writing the early part of chapter 2 the MC is sent to an [undiscosed] location by Rikton (your ultimate adversary). This location winds up being pivotal to the story moving forward (lots of important things are revealed in chapter 2).

Originally, I planned the MC to travel to this location alone, but that requires lots of monologue and exposition - not the most involving way to tell a story.

Instead, I pivoted and decided to use another character to develop dialog to explain what was happening. But rather than introduce another random throw away character, what if I took someone who'd been briefly introduced earlier in the story. That's part of the joy of building the airplane while you land it I guess - I have the overall story arc planned, and some of the more important character/character storylines sketched out... but the rest - the inbetweens are free form as I write the chapter. That's the fun, creative process that I enjoy most in writing.

So I looked back at some of the faces who'd already appeared at the Dollhosue club, I have an ulterior motive here - the club itself is a front that we learn about in this chapter, so bring a character from that venue into the next location builds continuity. Good so far.


I chose this image as a starting point, for no particular reason - I just liked her look. Taking that image prompt, some additional description and using the lineart method I described in the last post we evolve to this picture - which I thought was pretty good:

But, then I really let SD go, adding an additional LoRA to help the posing and doing another round of lineart to render I arrived at this picture - which I think is just achingly beautiful:

This is the reason I work in SD, when a workflow leads you to something breathtaking... it's the moment I believe it's not all slop being generated. (yes, the picture still needs some touch up and clean up).

With her look established I started writing more and more of the scene, and having SD render the different expressions I wanted her to convey. And the more I wrote, the more the scene and character evolved. I continued to add depth to her back story and realized I don't want her to be a one-off. I want her to round out the cast of LI's, maybe not as a main LI, but someone who I can use to drive one aspect of the story (conflict) forward.

So make her a full fledged character; she needs a name - Gail. And a LoRA... I prototyped one based on the range of expressions I rendered for the scene in chapter 2 - it's a bit limited, but a starting point I can use to render more and rebuild. And a back story, which I think I've got a good start on.

The upside's are - this extra character development is in the course of building out the chapter 2 main storyline - so I needed to write a bunch of this material anyway. And... using Gail I added an additional sex scene to the chapter. It's optional, but I think I've got a good mechanism in place to get to it and intertwine it in the story being told. Before this happy accident I was looking at chapter two only having a single sex scene -but actually being a ton of work, because depending on player choices the scene played would be one of 4 different alternatives. Now there are two scenes in the chapter, which I'm happier with as far as quality of content goes.

But... that leads to another issue, and one I'd actually like feedback on - if you're reading this and you've made it this far, please let me know what you think:

• I had originally planned to release 4 chapters in a year (quarterly, with a 3 month development cycle), each comprising 2 weeks of story time.
• Looking at the story/chapter flowchart, I'm wondering if I should be aiming for 6 releases a year (2 month development cycle) comprising 1 week story time. Each chapter will still be 2 weeks, but I'll break up the development & release schedule into game weeks.

In my mind chapters are major story blocks, each starting with a quick review of the prior chapters major decisions/crisis points and ending with a summary/introspection -like I did at the end of Round 1.

There are (at this time, and I have no plans to add more... but plans always change) 6 (7 actually, Ash & Leigh are treated separately in the important scenes) LI characters (major and minor). So, I'm planning to keep 7 storylines moving forward. I'm conflicted.

Chapter 1's development got to be punishing at the end of the cycle... constant editing and image reworks are just... work. Not free flowing like the initial concept images and writing. I'm afraid that the amount of material and editing needed for the 2 week (3 month) cycle is going to be too much for me to manage and stay properly motivated and engaged.

I'm leaning towards the 1 week (2 month) cycle, which I hope will also be more rewarding for readers: more frequent releases and will force me to focus on good content and not reusing the "my life is a grind - work, train, sleep" pattern. Just realize the downside will be less content (and sex scene) in a release - I will always plan to have one significant storyline event (and potential sex scene) per release and frequently those scenes will have different variants depending on the player decisions: case in point - in Chapter 1 the final Ash sex scene was different depending on if the player chose the BDSM or non-BDSM route (did anyone notice?), or if you entirely declined to have sex with Ash at that junction- with different outcomes in Ash's personality and the final scene, NVM if you had declined to even enter a relationship with Ashleigh.

Anyway - looking forward to reading opinions on this. Meanwhile, I'm fully engaged in writing the first week of chapter 2.

As always, thanks for reading,
Regards,
-Insomniac

 

[1nsomniac22]

Churn 'n Burn

TL;DR - spent some money, fell a bit behind schedule

This week hasn't been great for writing/progress on the story. IRL keeps getting in the way. I still haven't completed the first draft of the first week. It's outlined, but I haven't had as much time as I'd like for writing. I still have 3 days of content unwritten, including one major event with one of the LI's.

I did figure out one additional aspect of Elizabeth's backstory while writing another scene with her. In keeping with the theme of everyone being broken in some way I figured out how such a perceptive person wound up working in a strip club. Likewise, I've figured out how to close a story gap from the prologue with respect to Ashleigh and the drug buy. It doesn't quite put a bow on it, but I think the overall story arc and your motivations will be tighter and more compelling by the end of chapter 2.

On the other hand, I did buy a referbed RTX 4080 to put in an old chassis I had sitting idle. The box has less RAM than my laptop, but the new vid card has 16 GB VRAM, double the 8 GB on the laptop, which makes SD super happy. I set up Invoke in server mode on the new(old) box and am thrilled with how quickly it's now generating sketches and at higher resolutions... so that's nice. However, I experienced some stability issues on the new(old) box that I never had on the laptop. Doing some benchmarking and testing to see if the card is a dud... It seems to be pretty stable at 90% power throttling... which is weird since the box has a 750W Corsair PS, not a cheap component (it's my old gaming rig). Then while doing troubleshooting I lost the M2 HDD. Everything is working now after a bunch of config inspection and driver cleanup - waiting to see if the system stability issue is resolved.

During setup of the new Invoke install I noticed there was an architectural trained basemodel that I'd never seen before. In early testing it's doing a really nice job on the backgrounds. I will probably use it in all future BG generations (instead of the base SD 1.0 model), and will identify and replace most of the BG's that shipped in Ch1. Yes, there will still a visual style difference between the BG's and the characters... that's kinda my style.

Sorry to the respondent of the poll - I appreciate your taking the time to answer, but for this release I'm going to try the split release of the first in game week of Chapter 2 within 2 months and the second week in the following 2 months.

And, lastly, in keeping with giving some hints of how the story is shaping up, here is a sketch from the LI event I'm excited to write. This is from the Lilly route. Without spoiling too much of the story - her favorite activity is dancing. A lot of her backstory will go into the how the why and her fall. Here are a few still rough sketches of her clubbing:

 

[1nsomniac22]

Quick Update this week.

I've finished drafting the first week of chapter 2 - the plan to split chapter 2 (and potentially beyond) into partial releases still seems reasonable. The amount of content is still shaping up to be consistent:

Prologue [Day 0 - Day 15]:
 1773 lines dialogue
 - 1076 prologue
 - 697 ashleigh

Round 1 [Day 15 - Day 29]:
3166 lines dialogue
  - 506 chapter
  - 1670 ashleigh
  - 744 destiny
  - 246 lilly

Round 2 [Day 30 - Day 44]:
week 1 [Day 30 - Day 37]
1499 lines dialogue
  - 709 chapter
  - 165 ashleigh
  - 205 destiny
  - 270 lilly
  - 150 gail

The amount of dialogue will remain consistent(ish) - I'm working the project structure to keep more of the single path dialogue in the chapter script and move the optional dialogues/events to the character scripts (I didn't do a great job in chapter 1, so it's a bit more spaghetti code than I like).

I got the issue with the new workstation resolved, just need to tear it down to bare metal and reinstall everything. I'm thinking it was either a driver conflict or I corrupted some part of the python env Invoke AI uses. It's been quite stable after the rebuild and able to render out about 2x to 3x the content in the same amount of time as my laptop. I'll be stressing it soon as I transition from writing to rendering.

Here's one additional quick sketch from Lilly's event:

 

[1nsomniac22]

Chapter 2 week 1 - transitioning from writing to editing

TL;DR - primary writing of chapter 2 first week is completed

This week I finished the first pass writing of chapter 2 week one (and the second week is outlined), which means I've switched over from writing to editing... which to my own annoyance can be more important than the writing pass. It's one thing to light a keyboard on fire writing content as fast as it comes to mind, but it's an entirely different skill to go back through and smooth everything out.

To use a sculpting analogy the writer sees the sculpture in a block of marble and roughs the form, but a good editor is what shapes the art. This process requires reading through a few times to make sure the story, dialogue, and word choice work... and it takes a lot of time.

I've been doing a little research about using LLM's as an editor. Now that I've got more capable hardware I could see spinning up an instance of Ollama with an editor model to try... but there are major limitations to address:
- the current pattern is to submit less than 300 tokens to an AI (about 225 words)
- would need to break up the manuscript into chunks
- the editor model(s) doesn't(don't) seem to have a good grasp on a chapter (or multiple chapters) because of memory limitations
- Ren'Py scripts aren't like normal manuscripts
- Grammar editor vs creative editor; grammar is "easy", creative is less so.

Maybe someday in the future I'll look into training an LLM specifically to understand Ren'Py scripting to help with text editing, grammar and (if the models get better), character consistency. The Ren'Py Analyzer Visual Code plugin gives me some ideas since that tool already handles text/character summarization within a renpy script. But... that's a future project.

Back to Sucker Punch - during editing is where I get serious about generating the final set of character CG's. I usually have a bunch of sketches I created while writing - either as storyboards for the scenes or as concept place holders that I include in the Ren'py script so I can see how things look in the editing pass. But, this is the stage where I lock in the images... and it could be the most frustrating. If the AI tool couldn't create an image to match my vision I'd frequently go into a doom loop where I couldn't continue the edit until I had a close enough image -I freely admit I have some compulsive behaviors. I've dealt with them for a long time, so I can recognize when I'm getting truly stuck and at that time I break out - either by putting a place holder in or working on something else for a while.

With the lineart workflow I'm using in InvokeAI I think I should be able to streamline this process a bit and avoid getting stuck. At least, that's my plan.
But, for now, I think I'm still on track to ship Chapter 2 week 1 late next month (US holidays may still push the timeline a bit).

Lastly, I've gone back and regenerated most of the BG's in the project. I think they're in a better place. Here are a few examples:

OLD	NEW

As always, thanks for reading,
Regards,
-insomniac

 

[1nsomniac22]

Chapter 2 week 1 - fully editing

TL;DR - fully editing now

Unfortunately, this week (as always), IRL got involved. Some work issues, some medical issues. Everyone's healthy now, thank God.

Regardless, been editing for nearly the whole week, which involved a little bit of rewrite and a whole bunch of image generation. Plugging away as I can, but sent the Gail LoRA back to the oven - except this time the oven is running with 16GB of RAM and a whole bunch more CUDA cores. What used to take days is now accomplished in hours. I'll be experimenting with upping the LoRA size (parameters) in the future and increasing the data set size now that I've got so much more horsepower to play with.

I'm pretty happy with how week one has shaped up... but, I did a bit more brainstorming for week two, and I'm really excited to get to writing it - I think that part is gonna be good! I know no one thinks their work is -meh- to be charitable. However, I want to ship something that tells an engaging story, and with each revision/edit/realization I always feel the story gets better, tighter, and more compelling.

Here's two nearly finished images from week 1 that take place with Gail. This is a return to the MC's story arc about his fighting career... in a sideways way:




Game mechanic stuff
Ever since the first release of the prologue, the MC had an attribute called corruption. I always considered this the players experience of events during the telling of the story. The way the story develops the MC begins a bit naive and if not innocent, focused on his ultimate goal at the expense of some real world experience. Over the course of the story the player will succumb to or resist some of the corrupting influences depending on the players choices - but there's always a degree of corruption that will happen during the course of the story.

The corruption score will be used (is used) for enabling some story choices.

After the events of chapter 1 I am introducing a new stat/mechanic starting at chapter 2: the player's sadism score. As described in the sex shop dialogue between Ash and the MC, sadism reflects the MC's getting pleasure from the pain or suffering of others. It's beyond just the corrupting influence of the main story and is currently controlled by the players choices to indulge Leigh (and Ash) in their machoistic tendencies. This score will be used to enable alternate sex experiences with LI's as/if the player's decisions take the MC down the path of more extreme machoistic behaviors (or their reactions to events in the story).

As I see it, when the MC has a low sadism score it's unlikely they'll try to engage in extreme sexual behaviors with a LI, and if they have a "high" sadism score they'll be less likely to be able to engage in more normal sex (normal sex will be less and less gratifying to a highly sadistic MC)... but I'm still working this out - for the time being, in chapter 2 - if you indulged Leigh/Ash's sub desire you will have the option to pursue more extreme encounters (in week 2, sorry). I see this mechanic coming in to play with the other LI's as they're not inherently "into it" like Ash & Leigh are.

As always, thanks for reading. Still on schedule.
Regards,
-insomniac