Open-weight image generation and editing model from Pruna AI, distilled from MiniMax's H3 video model to run in 8 steps.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
P-H3 Image (also written PH3 Image) is an image generation and editing model released by Pruna AI. According to YouTube creator coverage, it is adapted from the MiniMax H3 video model and distilled to run in 8 inference steps instead of 50. Creators report it is about 7x smaller than the base model and generates a 4MP image in roughly 2.6 seconds.
Reported capabilities include text-to-image generation, image editing, output up to 16 MP, and combining up to eight reference images. It also has a prompt enhancer. Creators highlight photorealism and handwriting rendering as strengths. One reviewer who tested the free Hugging Face space found quality well below ChatGPT's image model.
Weights are reported to be on Hugging Face in BF16, 8-bit and 4-bit versions. The full model is roughly 55 GB and the 4-bit version about 21 GB. Pruna also offers its performance models through an API and a self-hosted option. The provided pages do not confirm the license or pricing for this specific model.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Image generator/editor adapted from the Minimax H3 video model, 8 steps instead of 50, up to 16 MP, combines up to eight reference images and has a prompt enhancer. Weights on Hugging Face in BF16, 8-bit and 4-bit.
Open-weight image model distilled from MiniMax H3 video model. Weights ~55 GB with quantized versions (21 GB at 4-bit). Free Hugging Face space tested, quality well below ChatGPT.
Pruna AI turned the MiniMax H3 video model into an open-source image generation and editing model. It runs in 8 steps instead of 50, is about 7x smaller, and makes a 4MP image in 2.6s. It is strong at photorealism and handwriting.
Closest neighbours by meaning, not just by category.
Model · MiniMax
Open-weights MiniMax image generation and editing model producing up to ~16 MP images in 6 sampling steps.
Model · Hao AI Lab
Distilled MiniMax H3 reference-to-video-and-audio model from Hao AI Lab, generating in 8 steps instead of 50.
Model · MiniMax
Open-weight video generation model from MiniMax, part of its H3 family alongside H3 Image.
Model · Hao AI Lab
Open-weight 4-step sparse-distilled version of MiniMax H3 for fast text-to-video-and-audio generation on Blackwell GPUs.
Model · Pika Labs
Pika's open-weight video model post-trained on MiniMax H3, generating synchronized video and audio up to 4K.
API · fal
Fine-tuned, low-latency variant of MiniMax H3 video model served through fal's paid API.
