Free ComfyUI workflows, Runpod templates, and guides: 👉 <a target="_blank" rel="ugc" href="https://hearmemanai.com">hearmemanai.com</a>For more LoRAs and updates, Join my Discord: <a target="_blank" rel="ugc" href="https://discord.gg/ZVWVhT43GW">https://discord.gg/ZVWVhT43GW</a>You can deploy MiniMax on Runpod with my template: 👉 <a target="_blank" rel="ugc" href="https://get.runpod.io/minimax-template">https://get.runpod.io/minimax-template</a> 👉 <a target="_blank" rel="ugc" href="https://civarchive.com/articles/34593/deploying-minimax-h3-to-runpod-with-comfyui-workflows-and-auto-prompting">Deployment tutorial</a><a target="_blank" rel="ugc" href="https://civarchive.com/articles/34593/deploying-minimax-h3-to-runpod-with-comfyui-workflows-and-auto-prompting"> </a><hr />UPDATED RECOMMENDED SETTINGSSampler - Euler Scheduler - Simple Steps - 12 Shift - 6 <a target="_blank" rel="ugc" href="https://huggingface.co/lightx2v/Minimax-h3-Turbo/blob/main/minimax_h3_fl2v_turbo_8step_v1.0_768p_comfyui_bf16.safetensors">Turbo LoRA </a>- Strength 0.5<hr />What this LoRA tries to achieve is realism and flexibility.V2.5 is not perfect, but it's a major step in the right direction. This LoRA was trained on close to 1000 videos with all kinds of acts ranging from penetration, handjobs, blowjobs, fingering, dildo riding, cumshots and it's pretty versatile.I deliberately added some examples that are not perfect to show that this LoRA is not a plug and play solution that always works and I encourage you to experiment with different prompting techniques and LoRA mixing.I am already working on V3 and happy to hear any feedback you might have!

HMNSFW - AIO Sex LoRA

15x the dataset size (close to 1000 videos) trained for over 40 hours. I was planning to release this as a V3, but I am still not fully happy with it, still this is a major step in the right direction, I am still refining this LoRA and expecting to release more versions in the futureYour prompt is EVERYTHING, a shitty prompt will get you a shitty result, use the provided examples from the embedded workflow as anchors. This version does I2V very well with coherent anatomy and motion and nails most of the T2V videos I thrown at it with the occasional bad seed.I recommend the strength to be anywhere from 0.5 - 0.9

Files with SHA256: a07732a84fd733085eb5d910f602f918fa7a3658117116927e4329f5951a9d2d