Free ComfyUI workflows, Runpod templates, and guides:
👉 hearmemanai.com
For more LoRAs and updates, Join my Discord:
https://discord.gg/ZVWVhT43GW
You can deploy MiniMax on Runpod with my template:
Runpod MiniMax template
Deployment tutorial
RECOMMENDED SETTINGS
Sampler: Euler
Scheduler: Simple
Steps: 8
Shift: 6
Turbo LoRA: strength 0.85
This LoRA was born from experimenting with training a V3 of the HMNSFW LoRA, mainly testing different training settings and seeing what's possible.
It came out pretty well so decided to post it.
This is a focused male POV LoRA for missionary and doggy scenes on MiniMax H3. It was trained on 26 hand-picked videos.
There is no trigger word. Write the prompt normally.
Description
Anuses appear more often, needs to be prompted (see examples)
Tried to make higher quality outputs with this one
FAQ
Comments (15)
Hello, I haven't been posting comments for a long time due to lack of time, but I'm taking the trouble to do so here just to tell you: your Minimax LoRAs are excellent, really excellent. The V1 had small color change issues sometimes, but changing the seed often worked well. It also had an excellent understanding of the prompt. I hope you'll continue to offer us wonderful LoRAs, thanks again!
Thanks man, appreciate you
Do these work with anime style images? Has anyone tested it?
thank you
Did you train with all videos or a combo?
videos
I always loved your loras and the workflow. All the workflows are gold standard for Minimax.
Question: How do you write the prompt in such details that actually generate what you prompted? I create in such details too (thanks to your workflow) but it's kinda hit or miss for some reason.
Keep up the good work!
thanks buddy.
Obviously it's not always perfect and I cherry pick.
I use Opus 5.5 for the prompt writing
@HearmemanAI does opus 5.5 writes nsfw prompt ??????
@TaiLong All Anthropic models write whatever you ask them to
Thank you. Appreciate your dedications to this community. :)
@HearmemanAI how ?? all of them gives lectures about restrictions regarding it 🫤 well not sure if you use some exceptional jailbreak [ which I don't think would work with Claude ] 🤔
@TaiLong Claude app / terminal / openrouter
You just have to slide it in during a conversation.
If you're going to straight up ask for a prompt for genitalia or something like that you'll get blocked
you just gotta make sure the scene and character descriptions are outside of the action and dialogue. generally works best when it goes "setting" -> "characters" -> "actions" +> "dialogue".
If they talk, then do things, then talk, it's easiest to have the descriptions between the dialogue pieces, rather than having to explain again. You can practically treat the "setting" and "character" pieces like you would Krea2. Like if you take a photo prompt, strip any references to photo and static movement, and then add actions at the end, it'll generally work T2V right out've the box.
It's generally only necessary to include a "general description" sentence or two at the beginning when the prompt is particularly complicated (with many shots) or with long settings and character descriptions. Sometimes when scenes change and stuff, it can help to quickly go through the whole list again, or at least the last action, to keep things constant (especially when doing V2V/extension).
I generally keep a huge text blurb at the bottom of all my videos to keep style and and stuff consistent. I find my prompts that are more just "me talking" do better than the prompts that an LLM fills with superfluous words, or even when I specifically try and follow the "H3 guide".
works real good. great to pair with other loras, it tends to by highest strength of my general stack. I just swapped from version 1 to version 1.5 (because I thought I done it already) and it was worthwhile catching. The video is less washed out now, which is a bit stylistic, but the movements and anatomy are more confident and formed. Feels like I just found a better strength on a turbo lora lol