Description

Image imported from Derpibooru.
Original: https://derpibooru.org/images/3503821
Metadata at the time of import:
Field Data
Favorites 9
Upvotes 17
Downvotes 3
Score 14
Comments 3
Uploader Ada
Original description below this line

Just another example from the new local text to video model HunyuanVideo. It knows quite a bit about pony and is 100% uncensored, even being able to do decent NSFW stuff. We are one lora away from unlimited, quick, high quality pony animations…
Here’s a guide on how to get it working, pretty sure the minimum is like 16GB VRAM for half decent generations:
https://www.reddit.com/r/StableDiffusion/comments/1h7hunp/how_to_run_hunyuanvideo_on_a_single_24gb_vram_card/
/ https://www.youtube.com/watch?v=wL87_UAmrDA
Hint: For low resolutions crank flow_shift, also raise embedded_guidance_scale. With 12 steps and a low resolution like this you can get gen times to under 30 seconds on a RTX 4090.

Comments

Syntax quick reference: **bold** *italic* ||hide text|| `code` __underline__ ~~strike~~ ^sup^ ~sub~

Detailed syntax guide

Background Pony #795D
I want to try to install this model on my computer using your instructions. Hopefully I can use it with my 3060 at 12 gigabytes, taking into account that the missing video memory will be borrowed from RAM. According to my biased estimates, with similar settings to yours, I can get the generation in about 8-10 minutes. Did you use the quantized version of the model? As far as I know, more than 24 gigabytes of video memory is needed

Imported from Derpibooru - Posted by VHS Studio
Background Pony #795D
@Background Pony #795D
Ты чекни слив с презентации Sora v2, вообще челюсть отпадёт. Выглядит пиздец как приятно и стабильно. Но за китайскими моделями всё равно преимущество в плане того что их обучают на АП контенте и никто им ничего не придьявит.

Imported from Derpibooru - Posted by VHS Studio