MiniMax H3 video model released with open weights, accessible via platforms
ByPulseAugur Editorial·[78 sources]·
MiniMax AI has officially released its new omni-modal generative system, MiniMax H3, which can produce video with synchronized stereo audio up to 2K resolution and 15 seconds in duration. The model is now publicly available on Hugging Face, with various configurations and weights provided for different use cases, including a LoRA for faster video generation. Despite its capabilities, the model's availability is restricted in certain regions like the US and EU due to evolving AI regulations, though it is being integrated into platforms like 秘塔AI, offering accessible, low-cost generation options for creators.
AI
IMPACT
Accelerates development and accessibility of high-fidelity video generation, potentially lowering barriers for creators and researchers.
RANK_REASON
Frontier-lab model release with open weights and system card.
MiniMax H3 × @fal 🎙️
This Thursday, join @OdinLovis, Creative Engineer at fal, Ethan Wei, AI Solutions Architect at MiniMax, and @VictorSuOrtiz, GTM Engineer at MiniMax for a live breakdown of how creators and developers are building with H3.
We’ll cover multimodal references, …
MiniMax H3 × @fal 🎙️
Tomorrow., join @OdinLovis, Creative Engineer at fal, Ethan Wei, AI Solutions Architect at MiniMax, and @VictorSuOrtiz, GTM Engineer at MiniMax for a live breakdown of how creators and developers are building with H3 on Fal
We’ll cover multimodal https://t.…
The MiniMax H3 open-source community is 🔥🔥🔥!
Looking forward to the MiniMax H3 benchmark, the MiniMax H3 robotics dataset, and seeing so many H3s around the world.
MiniMax has released H3 with open weights, combining text, image, video and audio in a single workflow. The 768p Base model runs on consumer GPUs, producing 15-second clips at 2K with native stereo sound. Community benchmarks show generation times of around 160 seconds for 5-seco…
<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1ve1mvh/minimaxh3_now_on_huggingface/"> <img alt="MiniMax-H3 now on huggingface" src="https://external-preview.redd.it/WyIvGqO4ylsfEfqo7vAaNtLhBYgl2urdqW5R5dAOYG4.png?width=640&crop=smart&auto=webp&…
<!-- SC_OFF --><div class="md"><p>I really did not expect this. I was thinking "wow that video generator looks so good, no way in hell that runs on my 7 years old gaming laptop"</p> <p>Guess what? I was wrong! I fired up the default workload for ref2va and first it inde…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vif7a0/minimax_h3_seems_to_handle_almost_anything_you/"> <img alt="Minimax H3 seems to handle almost anything you throw at it." src="https://external-preview.redd.it/bTJ0ZnRlem82MWloMXGV8tKPUcCpIhupkr9TD…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vhugqo/lightx2v_has_just_released_a_prompt_generator_for/"> <img alt="Lightx2v has just released a Prompt generator for the Minimax H3. Simply enter a short Prompt and let it work magic: no need more &qu…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vhpd51/ok_heres_my_entry_for_a_crossover_minimax_h3/"> <img alt="Ok, here's my entry for a crossover (Minimax H3)" src="https://external-preview.redd.it/NGd4OGk5bWdmdmhoMbfT-pnt6TfQKAHPBMsOkIkVfhMoy_ai_v…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vg4ky2/following_up_on_my_previous_post_and_yeah_im/"> <img alt="Following up on my previous post... and yeah, I'm honestly surprised by how well it held everything together. Still T2V only. Minimax H3 i…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vezce1/minimaxh3_weve_come_a_long_way_havent_we/"> <img alt="[MiniMax-H3] We've come a long way, haven't we?" src="https://external-preview.redd.it/YmZodHd2aHM1YWhoMQ9mflZ5qH0OephoI5owMLDi2F94WDtjBuJzC0D…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vex6os/minimax_h3_is_just_too_much_fun/"> <img alt="MiniMax H3 is just too much fun" src="https://external-preview.redd.it/MTlvbW55cmhuOWhoMfC16lhV5CvGKRu8n6bB0W7h-7cXkcWhTwz-w3jtTcRb.png?width=640&c…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1veofu3/minimax_h3_full_bf16_1920x1088_15s_single/"> <img alt="MiniMax H3 full bf16, 1920x1088, 15s, single generation, no offload. ~2hrs on a PRO 6000 Max-Q. It actually did the whole 3-scene prompt." sr…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1velx5r/minimax_h3_first_day_of_testing_mostly_just/"> <img alt="MiniMax H3, first day of testing. Mostly just having fun with it" src="https://external-preview.redd.it/bXdubnIzZHdkN2hoMfbO-cPHQdEDiE7dEub…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vegtac/minimax_h3_tips_and_tricks_and_what_i_experienced/"> <img alt="MiniMax H3 tips and tricks and what i experienced so far" src="https://external-preview.redd.it/tx5OS4bqaRRmp6ru2k8fSUiUvOuhT6aszTlpR…
<!-- SC_OFF --><div class="md"><p>Been playing around with this, and it could do literally all sort of things! Aside from t2v, i2v and r2v workflow.</p> <p>Like Wan22 and LTX23, it can be used as an Image Edit Model. You can literally set to 0.1 seconds (not sure how to set to on…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vecz21/minimax_h3_can_use_a_storyboard_as_a_visual/"> <img alt="MiniMax H3 can use a storyboard as a visual reference, this was my first one-shot test" src="https://external-preview.redd.it/bDU2ZjYzNHlyN…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1ve6lvr/minimax_h3_first_local_test_5_seconds_at_960540/"> <img alt="MiniMax H3 first local test: 5 seconds at 960×540 in 182 seconds on an RTX 4090 Laptop" src="https://external-preview.redd.it/MXpqYnE0c…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1ve442x/this_h3_model_is_for_real_thanks_minimax/"> <img alt="This H3 model is for real. Thanks Minimax." src="https://external-preview.redd.it/Y2trOTh5b3BmM2hoMYF8z-DNIe4lRaNSADgFoW3jr9qylNv0TPE-3PILr13p…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vdzkel/i_hope_this_is_not_the_case_for_minimax_h3/"> <img alt="I Hope This Is Not The Case For MiniMax H3" src="https://preview.redd.it/jfz80jckb2hh1.png?width=320&crop=smart&auto=webp&s=4e00…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1vdoqb9/repost_because_not_everyone_can_read_blurry/"> <img alt="Repost because not everyone can read blurry photos or long threads: MiniMax H3 is confirmed to be 33.1B parameters, but testers found out i…
<!-- SC_OFF --><div class="md"><p>This is from ComfyAnonymous in the Banodoco discord server</p> <p>"it's definitively going to be usable on a 3060, 832x480 124 frames takes less than 10 minutes right now on a 3060 + 32GB ram + nvme ssd using 8 bit weights and there's still …
<!-- SC_OFF --><div class="md"><p>I have been testing it via API. The benchmarks were not wrong when they claimed it was Seedance 2 quality.</p> <p>This model has flawlessly understood everything I've thrown at it. Even highly complex prompts like specifying exact times at which …
<!-- SC_OFF --><div class="md"><p><a href="https://x.com/MiniMax_AI/status/2083008095488516262">https://x.com/MiniMax_AI/status/2083008095488516262</a> (says it's coming in a few days)</p> <p>it can do text-to-image, and image editing</p> <p>everyone seems to be mostly hyped abou…
