Model
Trends
.ai
Model
Trends
.ai
the best open source ai models
Autoplay
Run Locally
minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
minimax_h3_turbo_v4_step600_ema_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
Download
minimax_h3_turbo_4step_ckpt850_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
Download
minimax_h3_turbo_4step_ema_ckpt500_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
Download
minimax_h3_turbo_4step_ema_ckpt850_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
Download
minimax_h3_turbo_4step_ckpt500_pruned_comfyui.safetensors
592 MB · SafeTensor · bf16
Download
minimax_h3_turbo_4step_comfy_pruned.safetensors
592 MB · SafeTensor · bf16
Download
MiniMax_H3.json
Other
Download
Downloads are served by Civitai. Some creators ask for a Civitai login first.
AD
Install / Run on Graydient
Requires membership
<larryvrh-h3>
About This Model
MINIMAX H3 · LoRAs
Created by
tsolful
on Civitai
by tsolful - Minimax H3 Turbo Loras larryvrh_4step_pruned
Creator notes
taomate_fl2va_3step_ema uploaded Current best settings: 4-6 steps, er_sde/euler, simple, Strength 0.75 minimax_h3_ref2v_turbo_8step_v1.0_768p_comfyui_bf16: Current best settings: 8-12 steps, er_sde/res_multistep, simple, Strength 1.0 lightx2v_4step_768p_v1.1+ v1.2: Current best settings: 5-8 steps, er_sde/res_multistep, simple. Strength 1.0 larryvrh_v4_600_ema: Current best settings: 6-8 steps, er_sde/res_multistep, simple. Strength 1.0 All examples are generated with 1.0 strength I did not train, distill, or create the original Turbo LoRA weights. This model page provides only modified compatibility versions intended to let the remaining compatible LoRA adapters load with ComfyUI's built-in LoRA loader when using the pruned/curve-form model. Full credit for the original h3_turbo_4step_pruned belongs to . And for the pruned comfy-compatible version Full credit for the original lightx2v_4step_turbo_v0.1 belongs to the team and for the comfy-compatible version Full credit for the original taomate_fl2va_3step_ema belongs to the team. And for the comfy-compatible version I'll be uploading different Minimax H3 Turbo LoRAs here and crediting the original creators
21,101 downloads · 826 likes on Civitai
SHA256
4F95AFB2AB503EAC876AE5571F4D78B0E0FD0198B626A89F4B7727018ED198A9
ModelTrends.ai Model ID
#26894
Flag content
Example renders
Prompts and settings shared by the people who made these renders on Civitai. Pick a render to see how it was made.
Settings
Steps
10
Size
864x1440
Prompt
https://civitai.red/images/138692349
Negative prompt
turbo lora (beta) strength 1.75, 3/7 steps + VFI
Settings
Size
768x1280
Prompt
integrated_multimodal_description: [Shot 1] Vertical first-person smartphone footage on a bright modern shopping street in daylight. The woman's right hand enters from the lower right holding a small imaginary doodle-conversion scanner: an obviously fictional 2D hand-drawn prop with an irregular black outline, white paper texture, cyan and pale-pink marker accents, and one simple circular energy gauge. The scanner remains identical in size, color, and shape throughout. The 1x main phone camera frames a white vending machine against a pale wall. The operator clicks the scanner trigger; a short cyan hand-drawn ripple and a few black ink dots touch the machine. A loose black-line targeting frame with cyan circles locks on for 0.8 seconds, then only the vending machine becomes a rough white-paper 2D doodle while its position, scale, perspective, wall contact, and drink slots remain unchanged. [Shot 2] At 00:03.750, the camera cuts to a sun-dappled sidewalk and tracks a passerby's transparent umbrella from behind. The scanner locks onto only the moving umbrella, preserving its angle, occlusion, and motion with the real female walker. After 0.7 seconds, the umbrella becomes a semi-transparent white-paper doodle with trembling black edges and cyan-pink hatch marks; the person, hand, clothing, and street remain live action. [Shot 3] At 00:07.500, the camera cuts to a parked city bicycle beside plants, paving, and a low rail. While walking, the operator points the scanner at the bicycle. Cyan lines travel along the frame for one second, then the bicycle becomes a cute static doodle bicycle with a white frame, uneven black outline, cyan basket, and pale-pink fenders. The wheels, chain, handlebar, ground contacts, gravity, and perspective remain physically exact; only the drawn outlines gently boil frame by frame. Keep every non-targeted surface and person live action. Use uneven black outlines, white paper fill, colored-pencil and marker hatching, ink specks, slightly overflowing color blocks, and subtle boiling lines. Use bright clean cyan-and-pink pastel color grading, modest smartphone HDR, soft digital sharpening, slight walking bounce, delayed tracking pans, small hand tremor, and minor auto-exposure shifts. Do not show a realistic firearm, bullets, explosions, injury, fire, text, subtitles, logos, watermarks, extra fingers, warped hands, layer drift, fisheye distortion, heavy depth of field, or excessive glow. overall_soundscape: Quiet pedestrian ambience, soft footsteps, a small trigger click, gentle electronic lock-on tones, marker-on-paper strokes, and tiny ink-dot pops accompany the live street sounds. non_diegetic_music: N/A
Negative prompt
https://www.seaart.ai/blog/minimax-h3-prompt
Settings
Size
1248x832
Prompt
1
Settings
Size
832x1248
Prompt
1
Settings
Steps
6
Size
768x1280
Prompt
https://civitai.com/images/139448487
Negative prompt
low-res 4 / 2 steps
Settings
Sampler
Euler
Steps
6
Guidance
1
Size
1536x2304
Prompt
For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced. integrated_multimodal_description: [Shot 1] Cinematic, high-quality two-dimensional animated movie with pastel colors. Tracking shot. At 00:00.000, the camera pushes in towards the cityscape slowly. A large futuristic airship with two rear-facing ducted fan engines passes by overhead, well clear of the skyscraper in front of it. The airship turns to the left and climbs, flying off screen. overall_soundscape: The wooshing sound of the aircraft flying overhead and the rumbling of the engines. non_diegetic_music: A triumphant cinematic score with heavy brass instruments and an upbeat tempo, reminiscent of 1950s era city of the future videos.
Settings
Size
736x1120
Prompt
integrated_multimodal_description: [Shot 1] Live-action, cinematic ASMR style, extreme close-up of a 22-year-old Korean idol woman with soft black hair, wearing casual tube top, fair skin and glossy pink lips, seated at a simple wooden table under soft warm indoor lighting. Close to a white ASMR microphone on the table in front of her on the left. A red-and-white KFC paper bucket is next to her at the right. She lifts a crispy fried chicken drumstick toward her mouth with both hands. The camera holds a static tight shot on her lips and the chicken as she slowly bites into the golden skin, pulling a strand of meat free with her teeth. Oil glistens on her lower lip. She chews deliberately with small open-mouth movements, then takes another slow bite while her tongue briefly appears to catch a crumb. She gets her mouth very close to the microphone at the left and whispers quietly: <d>[English, whisper] "Mmm, so good..."</d> overall_soundscape: Crisp crackling of fried chicken skin breaking under teeth, wet chewing and soft saliva sounds, quiet breathing through the nose, faint rustle of the paper bucket, and the subtle sticky sound of greasy fingers. non_diegetic_music: N/A
Settings
Sampler
Euler
Steps
6
Guidance
1
Size
1600x2176
Prompt
For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced. integrated_multimodal_description: [Shot 1] Cinematic, high-quality movie. Tracking shot. At 00:00.000, the camera moves backwards, focusing on the female Sister of Battle from the Warhammer 40K series in the center of the scene as she walks forward. Snow blows in front of the camera as the wind blows intensely, causing the woman's hair to whip in the wind. The camera arcs to the right at slow speed as she slowly walks forward through the deep snow. With a deep, low-pitched female voice, she says: <d>[English]The emperor has clearly forsaken this frozen hell.</d> At 00:05.000, she stops and looks to the left. The camera positions above her shoulder and focuses on the entrance to a dark, narrow cave in the side of a mountain. An echoing, guttural roar can be heard coming from the depths of the cave. [Shot 2] At 00:08.000, the camera hard cuts to the front of the woman as she raises the pistol in front of her. She walks forward towards the camera with narrowed eyes and a furrowed brow. overall_soundscape: The crunching sound of heavy footsteps in snow. The loud guttural animalistic roar of the unseen animal in the cave. non_diegetic_music: N/A
Badges for creators
Made this model? Put its standing on your Civitai page, your Hugging Face card, your README or your site. Each badge is a small picture that shows the standing on the day you copy it and never changes afterwards, however the rankings move later. The date is part of the badge.
larryvrh on ModelTrends.ai: #24 of 16,523 AI models
larryvrh on ModelTrends.ai: #1 of 150 MiniMax H3 models
larryvrh on ModelTrends.ai: #1 of 139 MiniMax H3 LoRAs
AI Models for Image, Video & Text Generation
larryvrh h3 is a LoRA for the MiniMax H3 family listed on ModelTrends.ai, a read-only catalog of open source AI image, video and text models. Compare it with other MiniMax H3 models, check its heat score to see how it is trending, and open it on PirateDiffusion or BitVector to try it.
Browse by Model Family
Browse all models