With the release of MiniMax H3’s open weights, creators and AI studios can now explore a new possibility: generating AI videos with native stereo audio directly on local hardware.
In testing conducted by MAXSUN Lab, the MAXSUN Intel Arc Pro Series successfully ran MiniMax H3 locally on a single GPU.
The MAXSUN Intel Arc Pro B70 32G Turbo generated a five-second video in just 153 seconds*—without relying on cloud APIs.
This brings advanced multimodal video generation directly to local workstations, giving creators greater control over their data, costs and production workflows.
*Tested using the default MiniMax H3 text-to-video workflow in ComfyUI at 864 × 480 resolution with a five-second output duration. Actual performance may vary depending on system configuration and workflow settings.
Tested Across Multiple Resolutions
MAXSUN Lab also tested MiniMax H3 across a range of output resolutions, from approximately 0.4 megapixels to 2.0 megapixels.
Both the MAXSUN Intel Arc Pro B70 and Intel Arc Pro B60 completed video-generation workloads reliably across the tested settings, demonstrating the practical performance of the Intel Arc Pro platform for local AI video creation.

Three Capabilities That Make MiniMax H3 Stand Out
MiniMax H3 goes beyond conventional text-to-video generation by combining multimodal understanding, native audio generation and advanced video editing within a unified model.
1. Unified Multimodal Understanding
MiniMax H3 can process text, images, video and audio within the same context.
Creators can provide product images, reference footage, voice-over audio and written instructions, allowing the model to interpret these materials together and generate a complete video.
This streamlined multimodal-input-to-video-output workflow can significantly reduce the number of tools and production steps required for advertising, social content and creative prototyping.
2. Native Stereo Audio Generation
Many video-generation models focus primarily on visuals, leaving audio production as a separate stage.
MiniMax H3 can generate video with native stereo audio—including dialogue, ambient sound and background music—as part of the same process.
This makes it particularly relevant for advertisements, Vlogs, music videos and other formats where sound and visuals need to work together from the beginning.
3. Advanced Video Editing and Control
MiniMax H3 also offers powerful video-editing capabilities, including video-to-video motion transfer and multimodal content editing.
These tools give creators more control over movement, visual direction and source materials, extending the model beyond simple prompt-based generation into a broader AI-assisted production workflow.
Bringing MiniMax H3 to Local Workstations
Running an advanced multimodal model locally requires substantial memory capacity, bandwidth and AI compute performance.
The MAXSUN Intel Arc Pro Series provides a scalable hardware foundation for these workloads.
MAXSUN Intel Arc Pro B70 32G Turbo
The MAXSUN Intel Arc Pro B70 32G Turbo features:
- 32GB of GDDR6 memory
- 32 Xe cores
- Up to 608GB/s memory bandwidth
- Up to 367 INT8 TOPS of AI performance
In MAXSUN Lab testing, the card completed a five-second MiniMax H3 text-to-video task in 153 seconds using the default ComfyUI workflow at 864 × 480 resolution.
MAXSUN Intel Arc Pro B60 Milestone 24G
For users seeking a different balance of performance and budget, the MAXSUN Intel Arc Pro B60 Milestone 24G offers:
- 24GB of GDDR6 memory
- 20 Xe cores
- Up to 456GB/s memory bandwidth
Its large dedicated memory makes it a practical option for developers, creators and studios building local generative-AI workflows.
Why Run AI Video Generation Locally?
Cloud-based AI video services are convenient, but they may also introduce limitations involving network availability, recurring API costs, rate limits and data privacy.
A locally deployed MiniMax H3 workstation addresses these concerns in several important ways.
Keep creative assets private
Product materials, unreleased campaigns, client footage and proprietary content can remain on the local workstation instead of being uploaded to a third-party cloud platform.
Reduce dependence on API availability
Local inference is not restricted by cloud-service quotas, interface changes or request limits. Creators can test prompts, adjust workflows and regenerate content as often as their projects require.
Build a more predictable production pipeline
For studios handling frequent generation tasks, local hardware can provide greater control over operating costs, system configuration and production schedules.
Customize the workflow
As an open-weight model, MiniMax H3 can be deployed through frameworks such as ComfyUI, Diffusers, SGLang and vLLM, giving developers greater flexibility to integrate it into their own production environments.
A Practical Platform for Local AI Video Creation
From AI advertising and rapid storyboard visualization to Vlog production and motion-transfer editing, MiniMax H3 opens new possibilities for multimodal content creation.
Combined with the substantial memory capacity and AI performance of the MAXSUN Intel Arc Pro Series, these capabilities can now move from remote cloud platforms to a creator’s own workstation.
One GPU. Local inference. Full creative control.
MAXSUN Intel Arc Pro and MiniMax H3 provide a practical foundation for the next generation of private, flexible and locally powered AI video workflows.



Deixe um comentário
Este site é protegido por hCaptcha e a Política de privacidade e os Termos de serviço do hCaptcha se aplicam.