Home Creators Posts Import Register Favorites Logout
haven't archived this post yet. have a subscription? use the importer!

Content

I've been hard at work implementing the archive prototype into my current setup. I spent a lot of time analyzing how the original naiv4vibe file stores data, so I can fully understand how its checksums and encodings work. But now that I got it covered, I can do a whole lot of interesting things. For instance, I've learned that every time you encode a vibe from the same image, a new extraction level is saved, and it can be retrieved if you change the setting back again.

That means that you can - in theory - save 400 different encodings into one single vibe file. It'd cost 800 Anlas, and wouldn't really be practical, but once all of them are encoded you wouldn't need to pay again.

Skipping hurdles

One of the problems I came across is that if you only have one encoding, it is fine enough, it shows as the default encoding once you upload it to NovelAi, it just works. If you want to change the extraction level, it generates you a new encoding from the image embedded in the file.

However, I discard all images and thumbnails when importing (copyright concerns and the risk of illegal or inappropriate embedded content). That means no new encodings can be made from my files, but the users of the file don't know of any other encodings that might be included either. That's where my new thumbnail generator comes in!

Every thumbnail that is imported to NovelAI comes with a little table showing all the encodings. It is not exact, by any means, but it allows you to see that 4.5 Curated has encodings at 0.5 extracted information as well as 0.7 and 0.9... or thereabouts. You can easily see inside NovelAI's image generator when there is an actual encoding, since then it will cost no Anlas. It's hard to present it in a good way, since I only got 100x100 pixels to show this info in.

The presentation might change, but it is what it looks like right now.

Wishlist: I want to include the creator somehow, but it is hard, seeing how there is no screen space left.

The Gallery

The gallery in the meantime is also getting its changes and additions. I'm repurposing the existing gallery code and expanding it to handle the Vibes, and if I can beat my own drum I must say I really enjoy what I got so far.

Hovering the gallery image allows you to see all the available encodings and the preview images created for them. The one with a green border is set as the default image (since it was the encoding settings that was saved with the vibe). You can also change the name of your encoding if you didn't like the original one, or if you forgot to name it in NovelAI before exporting. Clicking each and every one of the little previews let you see them in full size as well.

What's left?

There is a whole bunch of stuff left before I can show it to the public.

  • I need to hook up the favorite and voting system to this, which is harder than you might think since I need to connect two systems that don't really work the same, but still exist in the same space.

  • I need to make sure you can see them in your favorites viewer.

  • I need to add a filtering system to allow you to filter vibes on model.

  • I plan to double-check with NovelAI to ensure that this use of their API for generating previews aligns with their ToS. The way I read it, it requires user input (the upload), so it should be fine, but I want to be sure.

  • Testing, testing, testing, testing. So much testing...

  • Probably something more I forgot.

That's it for now

More will come later on when I start wrapping things up. Feel free to come with suggestions and feedback. Even if I'm weary of feature creeping, it'd be even worse not to include something critical I've overlooked.

Files

Previews only

Comments

No comments found for this post.