How to Add MiniMax H3 LoRAs on ClipMode

H3 LoRAs shape Animate clips — drone shots, expressions, outfits, and motion styles. Pick from the curated gallery or import your own H3-compatible LoRA.

Open My Models

What is a LoRA?

A LoRA (Low-Rank Adaptation) is a small add-on file that steers the AI toward a specific look or behavior — smoother motion, drone camera, anime style, better eyes, and so on. Your character LoRA from training is applied automatically. Style LoRAs are optional extras you pick in the generator.

When to use LoRAs

Motion & camera — fight scenes, walk cycles, arc shots, drone flyovers (video).
Quality & detail — eyes, skin, hands, film grain.
Concept / aesthetic — outfits, environments, art styles that are not in your base training set.

Skip LoRAs when a plain prompt already gives you what you need — too many stacked LoRAs can fight each other.

Where to find LoRAs

Copy the model page URL from one of these sites (not a raw .safetensors link):

Civitai
Large catalog; pick a version that matches your generator base model.
CivArchive
Mirror when a Civitai download is missing or locked.
Hugging Face
Many creators host .safetensors here — use the model repo page URL.

Filter for MiniMax H3 or Hailuo video LoRAs trained for the H3 engine.

How to import on ClipMode

Open your model → Animate → tap + LoRA / Style.
Tap Import LoRA (first tile) or My LoRAs.
Paste a MiniMax H3 LoRA page URL → Resolve → Add to library.
Select styles → generate your clip.

Custom imports require an active credit balance (paid feature). Pick LoRAs trained for MiniMax H3 — wrong base model gives weak or broken results.

Scale & triggers

Scale is how strongly the LoRA applies (0–1 typical). Start near the creator's default (~0.8). Lower (0.2–0.4) for subtle motion LoRAs; higher for strong style locks.

Trigger words are tokens the LoRA was trained on. When you add a LoRA, ClipMode can prepend them to your prompt — keep them in the prompt if results look weak.

Other ClipMode modes

Questions? Contact support.