MiniMax H3 Deployment Test - Runs smoothly with RTX 3060 12G, performance comparable to Seedance.

Recently, MiniMax H3 open sourceThe effect is comparable to Seedance, so I wanted to deploy it locally to play around with it, but the full-power BF16 version is too demanding on configuration, where am I supposed to get an H100!

However, I actually found a way to run it on my 3060 with 12GB of VRAM, which is to directly use the code from GitHub.open sourceThe ComfyUI project is a trimmed-down INT8 version.

Look at the results I got, they're really good!

hererecommendDownload ComfyUI Desktop directly from the ComfyUI official website. It's more in line with everyone's usage habits and the installation process is exactly the same as installing any other software. There are no obstacles during the download and installation process.

After installing ComfyUI Desktop and opening the template library, you will see three MiniMax H3 templates. Simply click to use them.

He will then remind us that a model is missing, and we can download the missing model.

After downloading, you need to restart and update before you can use it.

Once you've adjusted all the parameters, you can start generating the video. Actually, there aren't many parameters, just things like resolution; just fill them in.Prompt wordsThen it will begin generating.

Case 1 includes a built-in test case.

Prompt words：

Realistic live-action cinematic look, action movie trailer: practical film photography style, a post-rain dusk metropolis, anamorphic lens, shallow depth of field, film grain, city volumetric fog, flying-car traffic between the towers, restrained grading for a premium feel, powerful natural movement.

Scene overview: at dusk on a cluster of skyscrapers, the protagonist is being chased, sprinting and leaping across rooftops, jumping from one building's roof to the next with pursuers closing in behind. This is the escape sequence of an action movie trailer: every leap is life-or-death, thrilling and fluid.

Storyboard (each shot a separate scene, rapid cuts, all landing on the musical beats):

[0s-1.5s] Shot 1: high side angle: the protagonist sprinting at the roof edge, pursuers appearing in the rooftop doorway behind him, wind catching his coat.

[1s-2.5s] Shot 2: the protagonist leaps across the gap between buildings, body stretching mid-air, towers and flying-car light trails behind him, a slight slow-motion feel.

[2.5s-4s] Shot 3: he lands, rolls and rises, low-angle shot, tower shadows and fog behind him, he keeps running.

[4s-5s] Shot 4: freeze: the instant he hits the edge of the next roof and launches into the jump, silhouette, holding.

Camera: each shot its own angle, cuts clean and hard, no dissolves, a slight frame jitter on the jumps.

Audio: wind, rapid footsteps, city ambience, low score underneath, an accent hit on each leap, the score bursting at 4s, closing the last 1s.

No text, subtitles, logos or watermarks of any kind, no animation or cartoon rendering, no overly-CG look, keep the live-action texture.

This case has a resolution of 864×480 and lasts about 5 seconds. It took my 3060 8 minutes to run.

The cinematography has the feel of a live-action film, and the pacing is perfectly in sync with the music. The final cut is only 5 seconds long, but the few shots are very captivating.

As the protagonist dashes towards the edge of the rooftop, the pursuer emerges from a door behind him, his trench coat billowing in the wind, instantly establishing the chase. The first leap across the gap between buildings is the most beautiful shot in the entire video. The character is reduced to a silhouette, with the sunset, city fog, and light trails of the flying car behind him, clearly defining the height and distance of the buildings.

Case 2: Product video (960x544)

Prompt words：

Premium live-action macro product film. A black stainless-steel mechanical wristwatch with no brand name rests on a dark basalt pedestal inside a black studio.

Single continuous shot.

[0s-1.5s] Extreme macro view of the polished bezel and domed glass. A narrow warm light travels slowly across the metal edge.

[1.5s-3.8s] The camera performs a slow clockwise orbit while focus shifts from the front glass to the visible mechanical movement.

[3.8s-5s] The second hand moves smoothly. A soft cool rim light reveals the silhouette of the watch before the camera stops.

Camera: 100mm macro lens, slow precise movement, shallow depth of field, realistic focus breathing, natural optical reflections.

Lighting: warm narrow key light from camera left, soft blue rim light from behind, deep controlled shadows.

Audio: quiet mechanical ticking, one soft metallic click, low studio room tone.

No text, numbers, brand marks, hands, logos or watermarks. No warped watch case, duplicated crown or melting metal.

This case also took about 5 seconds, but it took 14 minutes to complete.

This is much more stable than the previous rooftop chase. The camera slowly moves from the side of the crown to the front, keeping the lugs, strap, and case intact, and the brushed and polished edges can withstand close scrutiny.

Warm light sweeps across the metal, then transitions to cool-toned contour lighting in the latter half, giving it the feel of a commercial watch advertisement. The entire video features no cuts or interruptions; the camera movement remains consistently fluid, which is quite remarkable.

Case 3 9:16 (480×864) Portrait

Prompt words：

Vertical 9:16 realistic live-action portrait film. After rain at night, an adult Chinese woman stands beneath the awning of a small convenience store. Warm indoor light falls across her face while the street behind her remains cool blue and softly out of focus.

[0s-2s] Medium close-up. She looks toward the rain, breathing naturally. A light breeze moves a few strands of hair.

[2s-4.5s] The camera slowly pushes closer. She turns toward the camera with a restrained, thoughtful expression.

[4.5s-6.5s] A passing car creates a brief moving reflection across the wet street and her face. She remains still as the camera settles.

Keep the same face, hairstyle, clothing and age throughout. Natural blinking, subtle breathing, stable eyes and teeth, no beauty-filter skin.

Camera: slow stabilized push-in, 50mm portrait lens, shallow depth of field.

Audio: soft rain, convenience-store refrigerator hum, one distant car passing, no dialogue and no music.

No text, subtitles, logos or watermarks.

This video is 6 seconds long and took 14 minutes to complete.

What I'm most satisfied with is the facial stability. The character's facial proportions, age, and hairstyle remain consistent as they turn from a full profile to a near-frontal view, and the eyes and lips don't distort during the turn. The wind blows the hair, and there's even a natural blink at the end. The camera slowly zooms in, the pacing is very restrained, like the few seconds before a character speaks in a film.

Case 4: Chinese Dialogue and Lip Movement

Prompt words：

Realistic live-action drama scene inside a small late-night noodle shop during heavy rain. Warm tungsten light, steamed windows, restrained cinematic color grading.

Single continuous medium shot. An elderly Chinese male shop owner places a bowl of hot noodles on the wooden counter in front of an off-camera customer.

[0s-1.8s] He walks into frame carrying the bowl. Steam rises naturally.

[1.8s-4.8s] He places the bowl down, looks toward the customer and says in Mandarin exactly: “Eat it while it’s hot, the rain will stop soon.”

[4.8s-8s] He gives a small friendly nod, wipes his hands on the apron and turns back toward the kitchen.

Natural Mandarin pronunciation, accurate lip synchronization, one male speaker only, no additional dialogue.

Camera: fixed medium shot with a very slow push-in, no cuts and no camera shake.

Audio: clear close male voice, bowl touching the wooden counter, boiling water from the kitchen, rain against the window, low room ambience. No background music.

No subtitles, text, logos or watermarks. Stable face, hands and clothing.

This video is 736×416, is 8 seconds long, and took 11 minutes to upload.

When the character holds the bowl, their fingers, the rim of the bowl, and the chopsticks remain relatively stable, and the action of putting the bowl down also conveys weight. There is no noticeable distortion in their face while walking, looking down, and speaking. The camera remains consistent throughout, only making slight zooms in, making the 8-second sequence very smooth.

The lip movements follow the sound, and the speech is pronounced accurately.Prompt wordsEat it while it's hot. The rain stopped soon. H3 mastered the timing of his speech and the boss's friendly tone.

This is a local deployment solution that allows a MiniMax H3 main model with a capacity of 60-70 GB to enter the workflow of a regular consumer-grade graphics card.

Project address: https://github.com/Comfy-Org/ComfyUI/

For lower-end configurations, an RTX 3060 12GB graphics card paired with 32GB of RAM is suitable for generating videos of around 5 seconds in length, at resolutions of 736×416 or 864×480.

recommendUsing 16GB of VRAM and 64GB of RAM will result in more stable 960×544 videos of 5-8 seconds; for high-resolution videos of 1344×768 and over 10 seconds, an RTX 4090 24GB or RTX 5090 32GB, along with 96-128GB of RAM, is recommended.

While 6GB and 8GB graphics cards might be able to boot up with model uninstallation, the waiting time and failure rate are relatively high. For most gamers, a 3060 12GB is already the starting point for seriously experiencing the MiniMax H3.

AI The video is now being processed into short videos, e-commerce ads, film previews, and product showcases.

For individual creators and small teams, visual effects are just the starting point. Generation costs, material privacy, parameter control, and modification efficiency also affect whether a tool can be used in the long term.

ComfyUI's MiniMax H3 INT8 solution brings camera control, character stabilization, Chinese dialogue, and native audio into the local workflow. Once the model is downloaded, creators can repeatedly modify it.Prompt wordsWith improved resolution, there's no need to upload internal materials, which also reduces the cost of per-call online services.

H3 can handle camera movement, lighting, character actions, and sound coordination, making it suitable for creating product trailers, short drama samples, advertising storyboards, and social media content. You can validate your ideas locally before deciding whether to proceed with formal filming, significantly reducing initial trial-and-error costs.

For individual creators, independent studios, and small content teams, this solution is sufficient to complete proof-of-concept and a small number of finished products.

When an RTX 3060 can generate a complete clip with camera movement, sound, and Chinese dialogue on the desktop, locally... AI The video has begun to enter the actual content production process.

Original link:A MiniMax H3 running on a 3060 graphics card has garnered 125K stars on GitHub.