Zhenzhen's AI Radar · MiniMax H3 · 2026-10-28 Evening Report
This issue features 2 new versions: Strongly Recommended 0 · Recommended 1 · Worth Watching 1 · Listed Only 0
Editor's Note This issue reviews 2 new H3 versions. Top pick: KinetiX Ground Fracture & Impact Feedback Slider: uses weight control to manage impact craters, cracks, and debris when slamming into the ground; includes four weight tiers and three-stage example prompts, upgraded from B to A. Loafy Style LoRA is a port of a furry illustration style imitating a named artist, with a nearly blank model card, downgraded from A to B.
Recommended (1)
[A] KinetiX · Ground Fracture & Impact Dynamics Slider · v1.0
LoRA · New Model · Author make676767 · Published 10-28 10:02 · Downloads 121
A weight slider LoRA that adds "heavy slam" physical feedback to H3 videos: positive weights cause ground slams and heavy weapon downward strikes to produce impact craters, radial cracks, flying debris, and a sense of impact, while negative weights result in smooth, non-destructive landings.
Upgraded from B to A: The four weight tiers and example prompts are specific and actionable, but only 1 visible examples are provided, which is not enough for an S rating.
Image
Reasons for Recommendation
- New capability: Continuously adjust the degree of ground destruction using a single weight, serving text-to-video shots involving ground slams, heavy stomps, and sudden stops in action/game cutscene contexts, without needing to repeatedly stack destruction descriptions in prompts
- The author provides four weight tiers: +3.0 ~ +4.0 for giant impact craters and debris storms, +1.5 ~ +2.5 for heavy weapon strikes (the author's recommended standard tier), +0.8 ~ +1.2 for fine cracks and dust from heavy footfalls, -1.0 ~ -3.0 for light landings with zero destruction
- Includes complete bilingual (Chinese/English) example prompts, organized according to H3's integrated_multimodal_description / overall_soundscape / non_diegetic_music three-part structure, ready to be copied and modified
- Free, license allows image generation, on-site rental, and rental; 11 hours after release, 121 downloads, 12 likes (indicating attention only)
How to Use
- Load KinetiX_dyn_fracture_impact_h3_v1.safetensors (3 MB · BF16) in the H3 workflow, no trigger word required, adjust tiers via weight: start with +1.5 ~ +2.5 for standard use
- Clearly describe the ground slam action, impact crater, cracks, debris, and camera movement in the prompt (the author's example uses a low-angle tracking shot with slight screen shake), and write impact and ground-cracking low-frequency sounds in the audio section
- Steps, CFG, and sampler are not specified by the author
Risks and Limitations
- Only 1 video examples are visible, 4 sensitive-level examples are hidden; the actual differences between weight tiers cannot be compared from the examples
- The author claims "zero color bias, character skin tone and lighting are unaffected," but there are no comparative examples to support this
- The file is only 3 MB, indicating a narrow scope of effect; the impact on non-ground-slam actions is not explained
Worth Watching (1)
[B] Loafy - Style Lora · Minimax
LoRA · Port · Author Foxntail · Published 10-28 15:59 · Downloads 38
An H3 version of the LoRA for artist Loafy's furry animal illustration style, primarily used for text-to-video; trigger word: loafy-style.
Downgraded from A to B: It has a trigger word and 3 examples, but it is essentially a style imitation of a named artist, with nearly blank documentation.
Image
Reasons for Recommendation
- Port: Previous versions existed for MOSS Image and Krea 2, but the H3 version's description only states "Mostly for text-to-video"
- Serves users who want to reproduce this artist's furry illustration style in H3 text-to-video; the model card does not describe style characteristics in text
- Includes 3 example videos, all with prompts; free, license allows on-site rental and rental; 5 hours after release, 38 downloads, 10 likes (indicating attention only)
How to Use
- Load loafy-style-minimax-h3-base-epoch-10.safetensors (296 MB · BF16) in the H3 text-to-video workflow, include the trigger word "loafy-style" in the prompt
- Strength, steps, and CFG are not specified by the author; refer to the prompts in the example videos
Risks and Limitations
- Training materials are works by the named artist Loafy, and the model card contains no authorization statement
- The model card is nearly empty, and the author does not describe the differences between the H3 version and other base model versions
- 4 sensitive-level examples are hidden, exceeding the number of visible examples
Trigger Word: loafy-style
Data: Civitai Public API (Base Model MiniMax H3), collection window 10-28 08:31 → 10-28 21:01 (Beijing Time). Recommendation reasons are editor-reviewed; download counts indicate attention only.