[Journey] $1.2 Million/year revenue within an 18 month period... Sharing the tech ;) you decide where it fits in your method.

timothywcrane said:

Which programming language/Toolset should you learn first?


I have never learned to program in any specific language. I have always approached programming as the skillset, and the language as merely a tool. No one becomes a "hammer specialist" before framing on the job....

I will have to hit w3c up to correct an unordered list in html and then bang out some obscure sed/awk/perl mess by heart into my term to complete something I know it does..... makes no sense if you look at it in the usual language fluency paradigm.

If you have 4 years to learn a specific code set for a boss you want to sign your checks for the next n years... then do NOT take my advice....

Fuck a language.... what's the problem? The language you learned on sololearn might not be the right tool for that problem.

Look at the word vomit in the above update... think you got time to know all that in the depth needed to be considered "fluent" enough to head off a corporate project on paper for each and every item listed?

Some truly are. Some aren't but have a team that does. I am not. I do not. So if you are new to the power button on your PC, have a masters in neural network technologies that you piloted, or anywhere in between... if you can cut and paste, you can code.

Forget about the language, you might learn French and get shipwrecked off digital Brasil. Quoting Demain, dès l’aube verbatim will do you no good, but one little word of Portuguese, Ayuda, could change everything.

Click to expand...

Ayuda is a Spanish word, You say Ajuda in Portuguese

Update... I'm addicted...

I have been creating art works for commercial print use for the last two months, 24/7, simply captcha-sitting and watching videos. I see amazing potential in the technology and am literally addicted. There is so much more I could be, and probably should be doing on this project, but honing in on the right settings for ML created works has seemingly been worth it. It takes anywhere from 15 minutes to over 6 hours to produce a single work, and finding the right balance of time/resource consumption has taken some fiddling.

So far I have created over 5000 high quality works of art in about 5 or so lore niches.

There are anywhere from 41 to 1841 layers in each art work in addition to the final render.
Currently all in 512x512px format.

The first transformation will be to upscale the resolution for 8ftx8ft stick-on POD wallpaper. I have already chosen a provider. This should be the highest resolution I will need. All else for POD and digitla creations should be lower. The wall will be set to render in Blender as actual size 1:1. The upload file to the producer will likely be much smaller.

The last will be to get legal IP documentation drawn up and create an NFT for each art work.

But there is a whole lot in between. I only have about 20 Gigs of storage left out of TBs locally, so the uploading and building will have to happen sooner rather than later...

timothywcrane said:

I got tire of that shit quick...
View attachment 201768


View attachment 201769

Click to expand...

Just be sure to rename image files before extraction or they will come out in a randomized mess... this is no fun if you hav a really detailed image with 1800 or more layers to reorder

timothywcrane said:

Images are only the beginning, but they are the simplest and have the widest range of immediate commercial, social and promotional/content usage...​

Click to expand...

Video and Audio ML integration with the studio will be the next order of business... and finally text/symbolic.

After overlooking the commercial Ai industry in a scrutinizing fashion for about a year now, from academic papers to Nvidia PR... imagery is the closest to being cut and paste for us laymen, but far enough along on the curve where most do not know you are not a graphics pro as of yet... (still a hidden tech trend).

Audio is next in line, with a good LTSTM strategy, any genre of music album can be generated with human assistance ASAP and even streaming audio is possible, though a little more glitchy. The ability to deep fake audio is almost cut and paste if you know a little audacity and aurdour for final edits.

Audio and visual replacement.. deep fakery is definitely ALMOST there. I can't wait to send an email to a customer in his/her own appearance offering hands off video ad creation (go ahead run away with it..)

Finally, video and CGI/gaming. I am undertaking this on two fronts... image to vid (traditional digital videocraft) with audio added, and ML live rendering, added to audio. The tech for the two is different, and most visual AIs we see are in dire need of a split in rendering. Those creepy AI videos with the Human form AI talking about the inferiority of humans take about 6 minutes to render the visuals with the Text responses for each 30 seconds of Youtube at best... This is not real time, and both the text response and the render are to blame.

On the gaming front I am leading out with DIY TCG wiki linked to artwork cards, and a natural upgrade to a Renpy based TCG once the lore is confirmed collaboratively. The move to a full 3d game is probably not the end goal, but using a gaming engine for AI driven scenic control for video/Meta env end use is something I have been playing with.

Enough of the blabbering.

Grab Krita, Gimp, Blender, Renpy or maybe even Godot (maybe wait for version 4 to settle) and a friendly AI and let's become media moguls.

and oh yeah... all of this can be split into collections or even single items, with commercial potential in the real world. Real value in the real world. Stick that in your NFT and host it on IPFS. Not just another fucking monkey... not a simple ephemeral coin investment... but an IP business for sale.

You can be anything from a cloth pattern whoiesaler , a music producer... hell anything text, image, audio, video, code,or node, in any mode. Full stack, from the original creative mind to the IP retainer.

1. Have AI/ML create
2. Human QA/edit
3. POD merch
4. Promote
5. Create community/Value
6. Wrap IP/Files on IPFS
7. Link IP (Intellectual Property rights) to NFT ownership and mint (with some sort of added visual media representation available, all IP/Files secured)
8. Automate and moderate community
9. Expand and explore third party opening. (public opening of my NFT marketplace etc)
10. Some secrets I ain't telling yet.

Just be sure to have a customer in mind before you create, an avatar at minimum, real names/title and contacts at best.

Bro I did not understand most of stuff but I liked what you are doing

could yoou explain little more ?

Prashurathore said:

could yoou explain little more ?

Click to expand...

sure. I post once an a while when I think a bout it, and often all at once... let me break it down.

The OP project is defunct. It was a political fund raising org. That is no more.

I've been sitting on my butt and making machine learning generated artwork...

This project is a full stack project that covers the complete traditional intellectual property journey of creative media and packs it into a single stack that can be controlled by the original creator from the beginning, as well as some machine learning and automating twists.

I am creating what I am calling a Creative Media Asset Generation Studio. I am creating works of text, imagery, audio, video, code, etc in many forms for both entertainment and commercial use, usually hand in hand.

I am using AI/ML with some human editing assistance to create everything from simple images to use on POD products, to music, to commercial "lazy deep fakes" (ie: you, the client is a video shy business owner and I create a video ad with YOU as the commanding star, and you do not even have to step in front of a camera).

You take that AI/ML created product and you create a full stack project for the IP (Intellectual property value)...

In my first use case I am creating an art collection, complete with an artist pseudonym.

1. I am going to POD the art...

Wallpaper
Wall Art
Home Textiles and Tapestry
Skate boards
Apparel
Prints
Personal print to own
"Seconds" Raw file sets for Artist/PODEditors with after-edit-PLR
etc...

2. I am going to create an artist studio website with the above mentioned software/section feature set in the table above.
This will allow me to manage/grow a community and content.
This is a great setup that can be used in two fashions or a combo of both.

ecoaching/elearning
collaborative media management

both are useful in this case...

I think Nextcloud/OnlyOffice is of a higher feature set and capability to offer "future value" to your community than a discord server... not to mention seccontrol concerns.

3. The goal is to gather a following for your art or other media... even a POD commercial success or two if you can...

4. Instead of trying to sell an NFT ape with a blockchain record in hopes of getting IP value out of it in the future... you create media that has a proven track record of IP value.... then make the NFT. The common sense proof of backed up valuation will come in handy as we go into the bear market and the excuse of rug pulls for fear mongering (it has already started).

5. Now if your shitty AI art is worth a measly $100/year in t-shirt sales/yr. Someone buys the IP rights for a fair market value of $100 to $1000. And that's if you only sell t-shirts.

6. You auto generate art at high velocity. Now I am not talking shitty untargeted creation, but focused material with a finite market in mind.

7. You have a plethora of client opps.. I am building in a collab balancing act for them all.

... in addition to POD purchasers and NFT IP Purchasers... there is a whole group of intersecting people in between that are available and funnels should be build for even if not the central target.

a. people who also want to create quality AI/ML generated creative art works., and learn to profit off of them.
b. Your genre specific audience has a complete stack of interaction...

I have one art collection that includes the photo art above.

I am targeting both Rebellious Punk/NuRock/Metal heads both 14 and 50 years old (two distinct target groups... think fathers and sons, not 14-50 year olds)
I will create Media in all media modes (images. text. audio, video, code) and target that audience. I will then sell the NFT to someone who wants that audience (my third target group)

The imagery as artwork and POD is the first step.

Before I hit hard on video offerings, I will use AI/ML to create accompanying music and other audio... and start with music videos using the layers of my imagery as a basis.

Think Dadabots but not live streamed. I will use more human editing and less of a disjointed metal style. For this I will have to create my own neural network and use LTSTM bursts with heavy human interaction and editing in audacity/Ardour. I want to make sure that I have the bulk of the imagery work done and Quality selections made before I endeavor on to this stage.

After Audio/Video additions, I will move on to CGI/Meta imagery...

Not all artworks will get this far.... just the socially/financially profitable ones...

I have a wiki to integrate with Art Trading cards, allowing anyone to create TCG based on the artwork , complete with custom POD cards with MY art for YOUR game.

There are tons of avenues, and as a fan of many income streams, I try to spread the wealth or lose it, so to say.

I guess that in a nutshell... you make media using machine learning and use it for everything everyone else is using traditional media creation methods to make a profit... and automate it.

Then you have a small business based on media with a proven income potential... you then wrap that IP up in an NFT smart contract,

Leave publicly viewable a few media bells and whistles for NFT showoffs and save ALL the IP associated files in a secure format on the Interplanetary File System, with the NFT and copy of contract and file hashes on the blockchain (I haven't decided on Ethereum or Polygon yet).

You should have your own IPFS node, so that you can "pin" the files to ensure "you ain't no rug pull". With a little JavaScript you can then be your own pinata service if you with (just to let you know). well worth a few hundred gigs of data.

It is obviously not a one post project and I post too few and far between... direct and specific questions would probably be best, if you can dig them out of the dumps above.

Thanks.

fresh off the Press... almost done rendering...



I have other collections based on chocolate candies or flowers... This is a generalized GAN based on CLIP...
So it has a wider range of uses than say FlikrFaces etc. and does a pretty good job IF you take the time to tune it for your needs.
I am also a fan of the sFlickr GAN modelset and usage. DuckDuckGo is your friend.

timothywcrane said:

fresh off the Press... almost done rendering...


I even have most of the site dev'd up locally... but I am just sooo addicted to the artwork creation... it a new artwork every time... even the mistakes are interesting and content worthy in parts... I am always amazed... Some aspects I have been using over this year are brand new but are already in commercial projects, even open source plugins for GIMP ( such as erase and replace functions AKA inpainting ) . Nvidia has led the way... but has been license restrictive to us looking to profit from our creativity and artistic ability, keeping it academic except for their own product usage. You can feed their AI but not profit from it.Building your own models for some generative art has been the only option other than Nvidia, but I guess that keeps the value high, if it can kept from Nvidia monopolization... think Microsoft in the early PC days with their installation bundling and customer service licensing model, but on Netscape vs Explorer anti-trust on steroids levels. Support open source AI nd especially ML modelling while you can.

Click to expand...

timothywcrane said:

For a good youtube channel on the latrest in ML/AI related to imagery and video generation and lighting, try "two minute papers".

Click to expand...

Even if ometimes the results are less than acceptable as is... Human assistance IS needed.

This is the life...or is it?

Just think... Some kid in Iowa could be getting a flabbergast look from their mom as they refuse to get out of bed in a bedspread that looks like this...

Trading card shit demo at the idea stage...

Very well explained and great info, thank you

a 2 minute hoodie makeover... shit edits just to fill it out...

An 8'x8' full size transparent image with an sRGB profile and at 300dpi will cost you 8-10 Gigs Memory in Gimp (and thus more to store after edits) for mock up for upscale resolution processing... will definitely need to get Nextcloud up or do some more disk shifting... I'm down to about 4TB...

timothywcrane said:

An 8'x8' full size transparent image with an sRGB profile and at 300dpi will cost you 8-10 Gigs Memory in Gimp (and thus more to store after edits) for mock up for upscale resolution processing... will definitely need to get Nextcloud up or do some more disk shifting... I'm down to about 4TB...

Click to expand...

Bro I don't have a single Idea what you are talking about but still following and see where it ends

For years we have been told to be careful when we edit images not to enlarge them at the expense of pixelation... that you can reduce images but not enlarge them without quality loss.... bullshit.
For a few years now we have been able to use AI to "increase resolution" by having ML "predict" and implant the missing resolution data when stretching a photo... or even just increase the
resolution of a standing image outright... Don't let anyone say you can't...

First... get a photo... I have a neural network (CLIP based GAN) create me a piece of artwork like above...

that gives you a 512x512 image (or optionally a 256)

then you can split that image into 64 - 64x64 images using GIMP.

You could simply enlarge your 512x512, but you might as well go for the lowest level to base your resolution advancement.




After splitting... head on over to Google Colab / 64 -> 512 and re-resolution each for 64 new full res 512s...

As you can see in the toolbar, we are currently sitting at 8 bit, good for consistency, not so much for color flexibility...
we will work on that with the ERSGAN output before we do any major editing or printing conversions.

Reading this entire thread in one go is like a fever dream, especially with the GAN images at the end. Good luck with it though haha