MiniMax H3 goes out of the box on Metaso AI, from 0.09 yuan per second

MiniMax H3, the video model that arrived with a surprise launch days ago, is now usable directly in a web browser. Metaso AI has added MiniMax H3 video generation today, so users can open a page, type a prompt, and generate video — no deployment, no GPU setup, no command line.

H3 is MiniMax's first open-source video model. After its release it topped the video editing category on the Artificial Analysis leaderboard, igniting the AI video community with capabilities including native dual-channel audio-video, 2K resolution, text rendering, and motion transfer.

Open-source models have always carried a deployment barrier: finding the repo, reading install docs, preparing a GPU environment, installing dependencies, and tuning parameters — a workflow that feels alien to content creators. Even after a successful setup, generating a 10-second video on a decent local GPU can take 20 to 40 minutes.

Metaso removes that barrier entirely. On its site, clicking video generation opens the MiniMax H3 page directly; prompt in and generate. Text-to-video, image-to-video, and multi-reference generation are all supported, and heavier users can plug into ComfyUI workflows for finer node control and process orchestration.

Price is the other weapon: 0.15 yuan per second for the 2K plan and 0.09 yuan per second for 768P, so a 5-second video costs less than half a yuan. In one hands-on test covered by QbitAI, a 15-second western-cowboy short took about three minutes from opening the page to finished video.

Notably, H3 bakes editing logic, typography, transitions, pacing, music, and visual effects into the model end to end, with precisely controllable multimodal editing. That makes it more than a better image generator — it compresses much of the post-production pipeline into the generation step.

In the industry's view, AI video is entering the era of prices measured in a few cents per second. When a top open-source video model appears in a mainstream product at near-zero barrier and a few cents per second, content creation shifts from mastering complex tools to simply having an idea.

The open questions are whether such low prices are sustainable, whether video generation will slide into a price war the way language models did, and how creators will reorganize their production workflows around this out-of-the-box capability.