# Dual Engine MiniMax-H3 + LTX-2.5

## Summary

One ComfyUI graph with a switch between a continuous `MiniMax-H3` chain and an `LTX-2.5` scene pass. It is a useful compare-and-contrast workflow for deciding whether a shot wants one long take or an edited cut structure.

## Model Family

- `minimax`
- `ltx`

## Status

- `imported-only`

## Source

- Workflow repo: <https://huggingface.co/joeygambino/Dual-Engine-H3-LTX25>
- Workflow file: <https://huggingface.co/joeygambino/Dual-Engine-H3-LTX25/resolve/main/workflows/DUAL_H3_LTX25.json?download=true>
- Original publisher: `joeygambino`
- Date imported: `2026-08-13`

## Developer notes

- The source README describes the graph as "one script, one switch".
- The `MiniMax-H3` lane is for a single continuous take.
- The `LTX-2.5` lane is for a scene that wants internal cuts.
- The source README points `LTX-2.5` users to `joeygambino/LTX-2.5-Quantized` for the smaller builds.

## Our notes

- This is the most interesting same-day hybrid workflow hit because it is not just another model dump.
- It is a better comparison tool than a first local run on an `RTX 3070 / 8 GB`; the full graph is more of a reference pack than a light test.
- If you want to judge H3 against LTX on grammar and continuity, this is the cleanest daily find.

## Required custom nodes

- `ComfyUI-H3-Multishot`
- `ComfyUI-Easy-Use`
- `ComfyUI-kjnodes`
- `ComfyUI-Impact-Pack`
- `ComfyUI-Ollama`
- `ComfyUI-GGUF` if you use the quantized LTX lane
- `RES4LYF`
- `WhatDreamsCost-ComfyUI`
- `comfyui-videohelpersuite`

## Required model files

- `Comfy-Org/MiniMax-H3`
- `joeygambino/LTX-2.5-Quantized` or an equivalent quantized `LTX-2.5` build for the second lane

## Notes

- Best treated as a benchmark pack for deciding which video grammar is worth pushing locally.
- The H3 lane is the more direct fit if you care about continuous takes.