MiniMax H3: China Tops AI Video Rankings
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
MiniMax H3: China Tops AI Video Rankings
China's MiniMax H3 video model is the first open model to reach the top of an artificial intelligence video ranking.
MiniMax has released the weights for the H3 model, marking the first time an open model has topped a video ranking. Artificial Analysis ranks H3 first in the video editing category, second in text-to-video, and third in image-to-video. This 33 billion parameter model processes text, images, video, and audio together, generating clips ranging from four to fifteen seconds with its stereo output. According to the model's specifications, a single prompt can include up to nine reference images, three video clips, and three audio clips.
However, two components remain closed. The 2K resolution module and H3-Context-IR, which translates prompts and reference material into a structured intermediate format, are not included. Running H3 locally in ComfyUI achieves a maximum of 768p, and users will need to manage context preparation themselves using the prompt guides published by MiniMax. However, the open weights allow for fine-tuning on custom sequences, characters, or specific visual styles. There is a licensing restriction: commercial use is only permitted for companies generating less than $20 million in revenue.
On the same day, ByteDance launched its closed model Seedance 2.5, which generates 30-second clips with integrated audio.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.