Apparently an open-source / free equivalent of GPT-4 will be released next year. Maybe a free, equally capable equivalent to MJ v5.2 / v6 (and Text-2-Daz3D) will be around by then too.*
^Open Interpreter is a pleasant surprise. Will have to experiment with a local install one day. I wonder if any of these AI programs can beta-read / critique fiction writing (editing works at a Masters / Doctorate level).
*Broken record time. Stable Diffusion may be one or half a version behind MJ, but it's catching up.
I donāt know what the current image size for MJ v5.2 is, however SDXL seems to handle ~1200x1200 (and 2x UpScale) without major issuesā¦except my SDXL copy runs out of Graphics Card spaceā¦
For an experiment, I tried a batch of three images in SDXL with 3x UpScale. Final image size 2700x3000.* I received one image before the graphics card ran out of memory. Still, the result was better than expected. But my graphics card doesnāt like me now.**
*
I also tried a 4x UpScale for the same image prompt (final image: 3600x4000), and received a warning / error message that at least 22GB of graphics memory is required for that task.
Some light Photoshop work required. Iām uncertain if the imageās errors are a result of SDXLās (and the Checkpointās) limitations, or my 12GB 3060ās limitations. I have no mouth, but I must scream ~ my GPU.
Oh, SDXL saves the final images as uncompressed .PNG files. This forum uploads the .PNG files as .JPEGs. So there might be a loss of displayed image quality / details.
Iām looking forward to the 3d modeling.
And hopefully the quality of video makers will improve.
I also hope that public opinion will quiet down a bit. There seems to be a consensus that AI images canāt be copyrighted so Iām guessing that will become standard going forward. Thatās fair. But if people could stop banning the use of them, that would be cool.
Text-to-3D AI software is making some impressive advances. But I doubt my current system could handle any local installments. An interesting work-around is to prompt Chat-GPT to create the Bender 3D code required to render a desired 3D object. Maybe next year there will be a (AI) text prompt / scene description to 3D object process for Daz3D, or one of the (quite expensive) enterprise 3D software packs.
A prominent concern I have with AI images and copyright is where, if, and how I could monetise the AI-generated images I created for my characters or my stories. Would be nice to receive some financial return on my investments (computer upgrades).
Oh, according to the System Info A1111 extension, any image prompt larger than 1800x1800 (up scaled) will max-out my graphic cardās memoryā¦and everything slows downā¦Unfortunately, the next largest AI-compatible graphics card (NVIDIA 4090 24GB) costs more than my whole system.
A young Karen Gavish roaming around her family's villa
Iām always surprised by how you manage to remain single ~ Polanski. What would a young, beautiful, intelligent woman want in a broken soldier turned assassin? ~ Zorik.
I donāt even want to fight over copyright. Iām feeling a lot less protective over my images nowadays because itās so much easier now to make something of good quality. Complex images still take some time but the basic ones are quick.
But I would like to be able to freely use them for commercial purposes.
For example, thereās an idea swirling in my head to record my book and then pair it with a series of images, like a montage as the scene unfolds. Like a graphic novel but in a video format and with full narration.
But I donāt want to invest my time in a project like that if itās going to receive negative feedback when people notice that Iām using AI.
Sigh. I donāt want to be boycotted before I even have a chance to make a dollar.
Draw some more and compete against them. Your marketing point is that they were drawn by a real human. I will actually buy one, if you make a digital coloring book to support you.
Apparently DLSS is the next big development in 3D animation and game / movie graphics. About 85% of the visible pixels of an image are created by AI, not rendered by 3D software, resulting in much faster image processing (without loss of picture clarity).
According to this video, past generation RTX graphics cards will be DLSS compatible. One lives in hope.
Deviant Art is also filling with AI generated images, some of questionable quality, and a few are for sale on the site (or via Patreon links). As DAās quality control seems to exist in name only, Iām reluctant to post any (AI) story-related images thereā¦or pre-Shutter Stock travel photosā¦Itās a flooded marketplace anyhoot.
Random thought. Perhaps create some postcard sets, character calendars, anime / manga style art-books with AI, and sell them on ebook or print-on-demand publishing websites. First examples I thought of were Blurb and Displate, and maybe Smashwords. J-List (NSFW content aplenty) or J-Box (a somewhat more family-friendly alternative) might be an option for Japanese style AI artworks or publications.
AI art character stickers are another option. If you can find a website that will create / host / sell custom sticker designs. 3D printing / etching of AI artworks (and selling their print files) would be a future project. I donāt know if the (free) AI technology is thereā¦yetā¦
