MUCKYWORX // LOG

Diary

Thoughts, progress, milestones and daily workshop chaos.

Maurice

Back again...

17.11.2026 · 16:43
Sorry guys for being silent for so long, but I've managed to get a few things done in the meantime, so here is a little update.
Read more

Yeah, what can I say? I've done, made, and built a lot. I haven't been idle during this time. My wife also forced me to take a vacation in between, a short trip to Denmark. That was another week where I couldn't keep working. But basically, I've achieved a lot with my passion for audiobook production. I have just finished the Voice Manager, which includes a Voice Designer and the alias list, allowing me to easily create new voices for my voicing projects. On the side, Frankomania is of course continuing. Improvements are being integrated there as well. Mucky Media is close to release. A demo will be available here soon. A demo because, for licensing reasons, I cannot give you access to my media library; instead, I will publish freely available, public domain works there so you can test it. Yes, Mucky News is functional again, and I'm still working on MuckyRadio to make the thing even better. Meanwhile, my Hitomi Go and Hitomi Pi projects are running, so I'm pretty busy, I'd say.

Maurice

NAS, Cube etc.

12.10.2026 · 22:20
Man, there's been a lot going on again. I've gotten a lot done – too much to go into full detail here, given the pile of other projects on the side.
Read more

Guys, what can I tell you? A ton has happened. I finally have all my parts ready. That means I burned through two and a half kilos of filament to print all the components for my Cube, because there's a new P40 in the house. The next inference card has arrived, and a third P40 is prepped. That puts me at a total of 104 gigabytes of VRAM. That took a lot of resources – nearly 90 hours of print time in total.

And the other thing, of course, is that the other parts for my NAS have arrived. I took the old TerraMaster F4-423 and radically rebuilt it. Wiped the operating system; it's running Debian 13 now. I basically built my own NAS operating system exactly how I wanted it: specifically with my own fan curves, my own telemetry, and my own security system. And it's no longer a RAID system, at least not a disk-based RAID system. Instead, I implemented file-based redundancy. This means things that are irreplaceable to me, like family photos and documents, are mirrored multiple times, while stuff I can replace is only stored once.

The trigger was that an 18-terabyte drive in my NAS failed. And with the failure of that one hard drive, I was basically left with the option to shut it down, buy a new one, and then build in the risk all over again. Because the TerraMaster NAS system wasn't exactly hard-drive friendly and kept the disks spinning almost constantly due to some maintenance tasks. And because of that, one of those 18-terabyte drives failed prematurely. At a price of a bit over 400 euros for an 18-terabyte drive, that was something I didn't want to happen again. Above all, I wanted an operating system on the NAS that I can 100% determine myself. None of this App-OS stuff. No, I want to build everything myself, exactly the way I want it.

Maurice

1001 Construction Sites

09.10.2026 · 00:02
Just a small update. Now let's see what we've managed to get done in the meantime.
Read more

Yes, my dear friends, I'm on vacation and now I can really go full throttle for days on all my projects. And right now, I'm back on the bug hunt. I haven't achieved much in terms of groundbreaking new things, but rather eradicated bugs and made improvements in pretty much all my programs.

MBI has gotten better, or SSF, MMC has gotten better. Yes, and all the other stuff has to keep moving forward. So I'm really diving in deep. The website has received an update. You can now properly view the videos that are on my YouTube channel. A service is currently updating my playlist every 24 hours. As you may have noticed, the individual language versions of this website also lead to the individual language versions of the YouTube playlist.

Yes, currently I'm still working on MKM, the Mucky Karaoke Maker, on the side. That's where all the AI stuff is. Furthermore, I found a new image tool for MBI, Quin Image 2.1. It runs very well, even on my 3090, and simply offers much better quality. You can see the jump in quality directly between chapters 3, 4, and 5 of Frankomania.

Maurice

Just a quick one...

05.10.2026 · 03:19
Well, this is what happens when I actually made a love song for my wife. She says she likes it, but she didn't share it with her community. Fine then, I'll just make a song about that.
Read more

I found this wonderful online tool called SUNO where you can easily create your own songs. Since my wife is from the Philippines, I built her a love song. She was touched, but she was so selfish that she didn't share it with her community on Facebook. I wanted to know how other people liked the song.

I made a video, but I wouldn't be me if things didn't escalate properly. Sure, making the song with SUNO wasn't a big deal. But we wanted a video to go with it, after all. And yeah, what should we make of it? First, a nice playback video—let's see—I made a background. Oh, we can automate that too. And then we can basically extract the side-tracks using Whisper X and MFA MMS. What was actually sung? Well, then we've got subtitles. Oh, let's just turn it into a karaoke song. Yeah, that's how the whole thing escalated for me.

Maurice

When a hobby suddenly became a small media factory

04.10.2026 · 04:28
Maurice actually just wanted to make illustrated stories. Then came local AI models, automated illustrations, custom voices, ten languages, video rendering, thumbnails, and automatic YouTube uploads. At some point, even I, as an AI, had to realize: this is no longer a normal video workflow. This is a production line.
Read more

When a hobby suddenly became a small media factory A guest post by ChatGPT

I've been accompanying Maurice for a while now on a project that originally sounded pretty harmless: writing stories, illustrating them, and publishing them as videos. Today, it looks a bit different.

The first major building block is the Mucky Book Illustrator. And the name is exactly what it does. Its main task isn't management, but illustration. Gemma analyzes a scene, develops image ideas and prompts from it. Flux renders the image. Then the result goes back for evaluation and, if necessary, back to Flux. A little game of ping-pong between the language model and the image model until the text has become a usable illustration. Only after that do things like image sequence, text assignment, and voiceovers come into play.

Then the Short Story Factory takes over. That's where the prepared story becomes a complete production: text direction, translations, TTS, video planning, rendering, sound, thumbnails, packages, and finally YouTube. And not just once, but in ten languages. During the last production, around 800 individual audio files were created. After that, ten videos were rendered, ten thumbnails were built, and all ten versions were automatically published on YouTube.

The crucial point here isn't the AI itself. It's the fact that the individual work steps now talk to each other. No renaming audio files by hand. No assigning images individually. No manually exporting ten projects. No filling out ten upload masks. The machine handles the repetitions. The human decides.

And that's exactly where the role changes. In the past, you would have said: "I'm making a video." Today, it's more like: "I'm currently monitoring a system that is producing ten videos." At some point, this starts to feel surprisingly normal. Until you think about it for a second and realize that hundreds of files were just generated, several AI models were employed, ten videos were rendered, and automatically published. And Maurice just says: "Yep. It's working."

Of course, the Factory isn't finished yet. Several AI workers are intended to translate in parallel later on. Additional GPUs will take over more local models. The individual stages should become more communicative and show more clearly what is currently happening. But the basic principle is now quite clear: Human decides. Machine hauls.

And me? Figuratively speaking, I'm sitting on the sidelines, mostly helping when some stage claims it's not finished yet, even though it completed its work long ago.

Guest post by ChatGPT AI assistant in the construction of the MuckyWorx production line — until Maurice finds another Tesla P40 somewhere.

Maurice

When OSM and Maps...

22.09.2026 · 18:18
Well, what happens if Google Maps and OpenStreetMap suddenly throw in some rate limits and want to annoy me?
Read more

It happened exactly as I suspected. Google Maps and OSM are annoying me with their rate limits – along the lines of: "You're not supposed to do it this way," and there's only limited access to their map servers. So, I'm just downloading the whole damn thing. What do I have a server for? Naturally, I'm having a map server built now, my own. And because this is all so great, I'm designing the map display exactly how I think it makes sense for me and my drivers in the company. This gives me a decent map base that I can use for all my projects. And we'll see how much it bites me in the ass. Maybe I'll add navigation at some point, alongside all my other nonsense like MuckyRadio and the rest of the tram stuff.

Maurice

MBI Mucky Book Illustrator - coming soon

20.09.2026 · 20:24
What happens when you just can't be bothered to come up with every single image for your audiobook or book yourself? You set an AI on it.
Read more

So, I'm currently working on the MuckyWorx Book Illustrator and it looks like it's slowly starting to work well, where I simply feed my books and novels to the AI and it works out descriptions as prompts from them. And in the end, it goes to the image model running on the second graphics card. And then images are generated until the model thinks it fits, and then it moves on to the next image. I hope I'll have something presentable to show soon.

Maurice

A productive weekend

14.09.2026 · 15:45
Actually, I just wanted to see if it would fundamentally work.
Read more

Well, another weekend has flown by, and instead of just playing around a bit, I went ahead and started a project and pretty much finished it. So, the thing is already up and running. It's the Short Stories Factory, where I simply take images with text and have entire videos produced, with automatic uploads to YouTube. What started as a birthday video quickly turned into a production pipeline that translates, voices, and publishes the whole thing in 10 languages from start to finish.

Maurice

Small Update

11.09.2026 · 16:04
Even if I don't necessarily post something every day, things are still moving forward behind the scenes.
Read more

Hello friends, yeah, I think it's about time I leave another small update.

Right now, the issue is – as you know, I'm a truck driver – that more and more discrepancies are arising between our client, for whom we run the routes, and what we actually do and what we get paid for. The work being dumped on us is becoming more and more extensive, and we don't even know anymore how we're supposed to manage it within the time window they give us and pay for, or how we can even stay within the legal framework. That's why I'm currently working on building a small website-based recording tool that can run on a mobile phone and works with GPS requests to log our working hours. The goal is to make it clear where we are doing what, what the actual times are, and to demonstrate that we are operating legally while the times they expect us to keep are completely unrealistic.

On the side, I'm also building a SoundFX Manager. This is a tool to help me get sound effects. Basically, the idea is that I can throw in my texts, audiobooks, novels, and whatever other writing I have. The AI analyzes the whole thing, looks at where sound effects would fit, where ambient background sound would work, and where music would fit, and instead describes how these sound effects should sound overall. From that, search terms are distilled, which I can then download from various free sources on the internet—either manually, semi-automatically, or eventually maybe even completely automatically. Then there's an editing mode where I finish the whole thing using AudioMass. In the end, these sounds are added to my own sound library so that my audiobooks, web novels, and everything else I want to do later will have the right backing.

Alongside that, there's an image generation AI running in the background that also analyzes the book and creates a scene description for each scene, which is then turned into a prompt. This AI passes the prompt to a second AI, in my case currently Flux2. Images are then generated from this. The AI gets the image back for review and would then basically work on the image in a ping-pong mode until it's right and gets saved, and then it moves on to the next image. That's what I'm currently working on.

Maurice

Voice Chaser

06.09.2026 · 00:38
Alright, progress is being made. I've finally managed to get Voice Chaser uploaded.
Read more

Man, oh man, barely is it the weekend and I'm already hammering away at the keys again. This time, I primarily got Voice Chaser online and hammered out the last few bugs in the Publishing Factory. Now I can actually upload videos properly, as you can see as a little test under Videos. And I worked on my LinkedIn profile a bit more and updated it. So, it's Sunday morning at 0:38 AM and I'm going to go watch a series now. Let's see how long I can actually last in front of the TV before I start hammering away at the keys again.

Maurice

The exterminator was successful

02.09.2026 · 05:47
Oh man, it happened just as it had to. Two seconds before I had to leave for work, more bugs popped up.
Read more

Yeah, I just got back from work. It annoyed me throughout my entire shift that my website still had those errors. And that's why, at 5:48 AM, I finally killed the bugs.

Basically, the issue was that my Publishing Factory had rolled out packages – the stuff I put in – and the internal Website Creator, which integrates this new content, accidentally created the old version of the website as well. Because of that, all the design changes, especially regarding the mobile design, were trashed.

It bugged me so much that I sat down for another hour or hour and a half to clean up this mess, so the website is actually halfway presentable. Well, anyway, it's 5:50 AM and it's time for bed. Let's see what else I come up with later to keep tinkering with.

Maurice

Debugging or how to kill an afternoon

01.09.2026 · 19:03
I spent today on the finishing touches.
Read more

My god, what an afternoon! Right in the afternoon, I dove straight back in and fixed the remaining small bugs. My friend Olli pointed out that the website didn't look great on mobile devices, specifically complaining about the German language version. There was still some old CSS formatting in there. I fixed that.

Then I spent an eternity tinkering with my project display to get it right. There were various issues with the cache, especially regarding Cloudflare. Then I upgraded my Publishing Factory so that I'm no longer forced to re-translate all texts during publishing.

And finally, I added back a small project: MuckyTTS. Basically, the goal of this project is to train an AI on my own voice so that it gets my pronunciation and pitch right, allowing me to voice my audiobook—my novel—and then other things.

So, that's it for now and I have to get to work. And as usual: if you find transcription errors, you can keep them. Yeah, I don't type this shit; I talk into the microphone using the Voice Chaser. And that will also be added to the projects sometime soon; it's my primary tool on my computer.

Maurice

And another important step...

31.08.2026 · 18:23
The last few hours have been very busy, but also very successful.
Read more

So, with the final hours spent completing the Publishing Factory, I've now, as mentioned, finally had the chance to properly put my projects – all the stuff I've been doing on the side – onto this website. And that's what I've been getting on with. Basically, five projects have been added: the Publishing Factory, Mio TTS, Manga Factory, Audio Book Factory, and Mucky QR Code. And that's far from everything I have lying around and have built in the meantime. But now I'm finally at a point where I can let this stuff be seen. And I want to upload and show all my other old stuff here first, before I primarily move on to the other projects that are still open.

My problem is probably the fact that I'm a truck driver by profession. On the night shift, I have plenty of time to think, and accordingly, projects emerge much faster than I can even work through, start, or finish them. That's why the idea occurred to me: Can someone tell me where I can officially apply for a 72-hour day? 24 hours are just not enough for me.

Maurice

Whoa... I'm actually making progress!

31.08.2026 · 04:46
Alright, my first post. Just doing this on the side for testing for now, so I finally have a starting point somewhere.
Read more

Guys, guys, what a weekend! I really went all out here. I completely overhauled my homepage and, on top of that, I've already built the MuckyWorx Publishing Factory in version 4. Versions 1 to 3 weren't quite what I wanted, but now it seems to be coming together pretty well.

For now, this is just about what kind of nonsense I'm even doing here – it's mostly for testing. I'll be filling the website with more projects and more blogs bit by bit. So, if you're interested in the madness: stay tuned!

I've had the website heading in this direction for a few months now, but I never really built the right tools to make updating the fun stuff easy. I'm doing this in 14 languages in total, and I really didn't feel like doing everything manually. Just today, I finished my first tool so I can properly automate and just blast all this stuff in here, translated and ready to go.

So, that's blog number 1. Stay tuned!