Minimax H3: EZ | Turbo . Optimal . RTX Upscale . LTX Refine — v3.0 - REMADE — V4.1 - LTX 2.5 Refiner

Use this ComfyUI workflow to make MiniMax H3 video clips from text prompts or image, audio, and video references, with optional H3 or RTX upscaling and LTX refinement.

Its modes include text-to-video, image-to-video, and reference-to-video, with a video-combine output.

The workflow lists 3 reference images, 2 reference audio items, and 1 reference video.

The publisher warns that release 0.31.0 will not work; verify which component that version refers to before changing versions.

Suggestion · not verified

Check that your models are compatible; the publisher says this workflow was made with INT8 in mind and recommends updating its nodes from GitHub.

Suggestion · not verified

The workflow lists these model files: 10Eros_v1.4_DMD_int8_convrot.safetensors, gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors, ltx-2.3-spatial-upscaler-x2-1.1.safetensors, ltx-2.5-22b-distilled-transformer-comfy-int8-convrot.safetensors, ltx-2.5-audio-vae-bf16.safet

It also lists ltx-2.5-video-vae-bf16.safetensors, LTX2.3_Crisp_Enhance.safetensors, minimax_h3_audio_vae_fp32.safetensors, minimax_h3_latent_upscaler_3d_bf16.safetensors, minimax_h3_video_vae_int8_convrot.safetensors, minimaxH3INT8INT4_fl2vaINT8Pruned.safetensors, qwen3vl_32b_min

For custom nodes, it lists comfyui-kjnodes and comfyui-videohelpersuite.

Turbo LoRAs and other booster settings are optional; the publisher says you can run without them.

For an LTX pass, choose either H3 or RTX upscaling; the publisher warns that using both first can make the video too large for LTX and exhaust memory.

Suggestion · not verified

The V4.1 notes say LTX 2.3 models were replaced by LTX 2.5, while the listed files still include LTX 2.3 names; verify those filenames against the workflow before downloading.

Suggestion · not verified