Generation: images and video with AI|INEMA.CLUB-PRO

Lesson 7 · 18 min

The open path: ComfyUI with LTX and Wan 2.2, VRAM and license

An open model runs on your machine, with no credit per generation and no need for a platform to stay online. In return, it asks for a graphics card with memory, free disk space and a read of the license. This lesson helps you decide whether your machine can run it and which model and template to use in ComfyUI.

Start the lesson
A dark workstation with a glass-sided computer showing a large graphics card and a monitor with a node editor connected by lines.

In this lesson you will

  1. 1What ComfyUI is and why it uses nodes
  2. 2How much VRAM and disk each model needs, per the official documentation
  3. 3Apache 2.0, community license and non-commercial: the difference
  4. 4The official Wan 2.2 and LTX-2.5 templates

Goal: decide whether your machine can run open video, pick a model and license, and record the official template you will use.

01From computer to open clip

Five decisions, in this order. Click each one.

Click each block in the diagram to see what it does.

If your machine can't run it, nothing is lost: the same still, the same prompt and the same list work for a closed generator.

02ComfyUI, models and licenses

Numbers from the official documentation, checked on 10/19/2026.

1. What ComfyUI is

A free program, with a window, that builds the generation as boxes connected by lines (nodes): load model, write prompt, load image, generate, save. Each box does one thing.

You don't have to build it from scratch: the official templates come ready, and the download-models button fetches the files that are missing.

03Template to copy

The sheet for your machine, which the page checks.

maquina.txt (template)
GPU: <your graphics card, or "none compatible">
VRAM: <the card's dedicated memory, in GB>
Free disk: <free GB on the disk where ComfyUI will go>
Model: <Wan 2.2 5B, Wan 2.2 14B, LTX-2.5... or "none, I stay with closed">
License: <Apache 2.0, LTX-2.x community license, non-commercial>
Template: <name of the official template you will open>
Use: <what you will generate on open and what stays on closed>

04Check your machine sheet

Paste the maquina.txt (or load the file). The page checks the seven fields and whether the license was cited.

maquina.txt

05Optional: check in the terminal

Only for people who already use a terminal. Download the course from GitHub and run it in its folder: the rules are the same as the page. For your own file, replace --exemplo with its name (maquina.txt); it exits 0 when it passes.

Terminal (in the course folder)
python3 conferir.py aberto --exemplo; echo "saída: $?"
python3 conferir.py aberto --errado; echo "saída: $?"
Example from the repository
$ python3 conferir.py aberto --exemplo; echo "saída: $?"
OK: ficha da máquina com 7 campos
saída: 0

06Practice

Without a compatible card, practice 1 still counts the same: recording "none" is also a decision.

Sheet · your machine and the model

15 min

Find out your machine's VRAM and free disk space. Pick the model that fits (or "none, I'm staying with closed"), read its license, and note the official template. Paste the maquina.txt into "Check your machine sheet" and download the file.

Done when: the page replies OK: machine sheet with 7 fields and the maquina.txt is in the project folder.

Optional · a clip in the open

45 min

If your machine can run it: install ComfyUI, open the image-to-video template for the model you chose, load a still from lesson 3, and use the motion prompt from lesson 4. Save it in 16:9 and check it in lesson 4.

Done when: the checker from lesson 4 replies OK: [s] s · [w]×[h] · 16:9 for the clip generated in ComfyUI.

07Quick quiz

Four questions. The explanation shows up after each answer.

Question 1 of 4

Your card has 8 GB of VRAM. According to the official documentation, which video model from the lesson fits?

08To keep

VRAM decidesWan 2.2 5B in 8 GB with offloading; LTX-2.5 asks for 32 GB or more and 100 GB of disk.
License readApache 2.0, community with a revenue cap, or non-commercial.
Official templateStart with the ready-made one; in LTX, turn off the prompt enhancer so you can use your own.

Going further

ComfyUI: downloadWan 2.2 on ComfyUI (official documentation)LTX on ComfyUI (official documentation)LTX-2.5 on Hugging Face (weights and license)
Previous lessonNext lesson