Tensor Works MiniMax I2V Forge — H3 Image-to-Video

Tensor Works MiniMax I2V Forge turns a still into MiniMax H3 video with native sound. Turbo LoRA stays on. One JSON, empty character slot, MysticXXX and Deepthroat baked in and off until you turn one on.

The still is cropped to the frame (not stretched). Ships 16:9 at 1.0 MP. Optional 48 fps is on the canvas but off until you enable it.

You do not rebuild ComfyUI, hunt custom nodes, or paste a GitHub token. Launch the Tensor Works MiniMax I2V RunPod template, optionally turn on a LoRA env var, load the workflow, drop a still, and generate.

WHAT YOU CAN DO

• Animate one still with MiniMax H3 image-to-video (first-frame / FL2VA)

• Opening frame is cropped to the selected aspect — no stretch

• Keep native H3 audio on the clip

• Official I2VA prompt fields already in the JSON (description, soundscape, music)

• Enable MysticXXX (0.28) or Deepthroat (1.17) from the Power LoRA node — not both

• Drop your own character LoRA into the empty first slot

• LightX2V turbo v1.1 768p at 4 steps already wired; leave UNET, text encoder, VAEs, and Turbo LoRA alone

• Optional 48 fps RIFE path on the canvas, off until you turn it on

WHO IT IS FOR

Creators who want MiniMax H3 image-to-video on a one-click cloud GPU, without maintaining a local ComfyUI install.

Works best with a clear still and a short English prompt that treats that still as Picture 1 / the opening frame.

WHAT YOU GET

• TensorWorks_MiniMax_I2V_Forge.json — MiniMax I2V with turbo, character slot, and both motion LoRAs

• TensorWorks_MiniMax_I2V_Forge_Setup_Guide.html — professional step-by-step setup guide

• MiniMax I2V RunPod template: https://console.runpod.io/hub/template/j7i55vzjnw?ref=wce04t8e

• Access to Tensor Works Discord support

WHAT YOU NEED

• A RunPod account and credits

• A GPU with 48 GB VRAM or more (recommended)

• A 200 GB+ persistent volume at /workspace (150 GB minimum)

• A still image to animate

• Set DOWNLOAD_LORA_MYSTICXXX=true and/or DOWNLOAD_LORA_DEEPTHROAT=true before first boot if you will turn those LoRAs on

RECOMMENDED HARDWARE

• Best: RTX PRO 6000 (96 GB), RTX 6000 Ada, A6000, L40S

• Volume: 200 GB+ persistent network volume at /workspace

• Lower VRAM cards may fail on 1.0 MP or longer clips — drop megapixels if you OOM

IMPORTANT NOTES

• This is a separate $9.99 product from Tensor Works MiniMax Forge (reference-to-video), Reference Forge (Z-Image), Krea Forge, and Krea Edit Forge

• RunPod usage is billed separately through your RunPod account

• First launch on a new volume downloads MiniMax H3 I2V (about 52 GB), then any LoRA you enabled

• Do not change the UNET, text encoder, VAEs, or Turbo LoRA

• Do not use the MiniMax Forge (R2V) template for this JSON

• Turn on only one motion LoRA at a time

SUPPORT

Questions or stuck on deploy? Join our Discord

https://discord.gg/HYDn58GKAX