V8 is a mix of the past v7 v6 and v5.
V8 mix B has better knowledge of characters and need less weighting to get colored skin for example. Not that it is always correct but it has a better idea of who a character is.
Either way both can be interesting, curious which one you prefer.
Rule of thumb for both hyper versions, if you get artifacts either our cfg is too high or your steps too low. Hyper versions are Euler and Euler a only most other samplers don't really work.
Version 6 seems to behave with most samplers, not sure if I would recommend Euler a however.
Removed 5.1 it was too buggy so I removed it.
Not sure if Clip merging is such a good idea especially with Pony models.
Did remove the T5 clip merge as it looks like it does not work, or only has an effect with A1111. If you are using Comfy UI there is no difference between v4 with or without T5.
The real version 9 of my fast Pony merge.
I decided to publish this under a new name as it will where I try to get the look I want (more then in the other one where I tried to keep everything Pony intact).
I don't think I broke anything Pony like yet...
Will leave the name however as "fast" as it is based on version 7.5 and you can get decent results with 10-12 steps.
Don't use Euler a, especially with version 2 (at least for me it causes issues)
DPM ++ samplers are fine
I hope nobody minds the prompts I have been borrowing to se how mine compares.
Description
non PCM version
FAQ
Comments (15)
This model works great and has come a long way from v1.
I love your work on this model and that you are trying different methods and constantly evolving it. The blend with T5 is especially fun for prompting.
Thanks, I can’t help it I like experimenting 😆
Glad you like it and thanks for the buzz.
Thx for your great work.
I am a bit interested in the version of V4 (T5xxl), actually, I've tested it and it do outperform the V4 without T5.
however, when I tried to perform the same merge process, the output weights are exactly the same with the clip model, the T5xxl merge seems not working.
BUT, the version of T5xxl do have difference between the version without T5xxl. Here is some difference I've found:
Weight difference:
first_stage_model.*
conditioner.embedders.0.transformer.text_model.embeddings.position_ids
Weight lack:
denoiser.sigmas, which is not in the version with T5xxl.
Just a bit interested in what exactly happened during the merge lol
Not really sure or if it is a bug with Comfy UI as I used that for the merge. Also it worked with 2 out of 3 checkpoints that I tried.
@TiwazM Whichever CLIP you stick into slot 1 will become the "base model." If this happens to be the slot with T5 in it, then it will save the T5.
Otherwise it will save whichever type of clip model. Just make sure the intended output type is in slot 1.
@Triple_Headed_Monkey Thanks, good to know.
V7 Vanilla? That was kinda......................fast?
Thx. And another tons of gigs to play with :-)
Of course it was ready to upload just had not done it and I needed to make a few pictures to test it a bit.
I am new to AI image generation and appreciate your efforts in creating this amazing checkpoint. I would like to ask if I have permission to sell the images generated from this model.
that is a tricky one to answer, as I used other checkpoints and even the Loras I trained are not trained on pictures I own the copyright too.
@TiwazM Ah I see, Thank u.
@isharamperera478746 maybe it is not an “issue” as it is all mixed up but I don’t know and I am not a lawyer.
@TiwazM ya I got it.
If you would like to sell the images, do your search for base models that have not been mixed. Also be sure to check the T&C for the model itself. I hope you enjoy using AI.
@Ruoar Thank u soooo much dear
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.