Get Started
The Lethal Stack

Mission 1 of 4

The Lethal Stack, Module 1: A face, a voice and a brain

Three free, open-source projects from our "Lethal" reel: one puts a face onto any video, one changes your voice live, and one gets you into AI models without an API key. This free module gives you the links, shows what each one really does, what people use them for, what your computer needs and the rules to stay on the right side of.

15 minBeginnerAll three projects are free and open sourceFree mission

What you'll do

The official link for each of the three tools, so you never download a fake copy
A plain-English picture of what each one does, with real output from our own test run
What people actually use them for, and the uses that will get you in trouble
  • What your computer needs for each one, and which to install first
Before you start
  • Nothing installed yet. This module is a read, not a setup
  • About 15 minutes
  • No coding experience
Your progress0/6
01/06Meet the three tools0 done
STEP 01 of 06~4 min

Meet the three tools

Know what each tool is and where its real home is before you install anything.

The reel called them a face, a voice and a brain. Here they are in the order you will set them up, easiest first, with the link to the original project. Always start from these links. Popular tools attract fake copies with malware inside, and the only safe source is the original repository on GitHub.

1. The brain: GPT4Free

github.com/xtekky/gpt4free ↗ (over 67,000 stars, GPL-3.0 licence). AI companies sell access to their models through an API, a paid connection that bills you for every word. GPT4Free is a free program that connects to AI services through other routes instead, so your own scripts and apps can use them without an API key. Its own page describes it as a project that "aggregates multiple accessible providers", and it comes with a chat app that runs on your computer, a Python library and a local API that other apps can plug into.

Our own run on 11 Nov 2026: three lines of Python, no account, no API key, and ChatGPT answered.

2. The face: FaceFusion

github.com/facefusion/facefusion ↗ (over 30,000 stars). You give it one photo of a face and a video. It finds the face in every frame of the video and replaces it with the one from your photo, matching the head angle, the lighting and the expression. It also does lip sync, face enhancement, age changes and background removal.

Left: the one photo we gave it. Middle: the original video. Right: FaceFusion's output. Both people are AI generated, nobody real.

3. The voice: Voice Changer

github.com/w-okada/voice-changer ↗ (over 21,000 stars). Its own description: "software that performs real-time voice conversion using AI". You talk into your microphone and a different voice comes out, with a delay short enough to hold a conversation. It plays voice models made with RVC ↗ (over 39,000 stars, MIT licence), a separate project whose own page says you can "train a good VC model with voice data <= 10 mins".

Checkpoint
You can say in one line what each of the three does, and you have the three GitHub links.
Stuck? Ask the community
STEP 02 of 06~3 min

What people actually use them for

See the real, legitimate jobs each tool does, so you know which one is worth your time.

GPT4Free

  • Learning to build with AI without a bill. Write your first chatbot, summariser or automation and pay nothing while you learn.
  • Prototypes. Test an idea against a real model before you decide it's worth paying for an API.
  • Trying models side by side. Ask the same question to several models from one place.
  • A local chat app that keeps your conversation history on your own computer.

FaceFusion

  • A consistent on-screen character. Faceless creators use one AI-generated face across every video, whoever films it.
  • Privacy. Appear on camera without showing your real face.
  • Film, comedy and adverts, with the permission of everyone whose face is used.
  • Fixing footage. The same app sharpens blurry faces and restores old video.

Voice Changer

  • Streaming and gaming. A character voice for a stream or a role-play server.
  • Privacy on calls. Speak without giving away your real voice.
  • Voice-overs. Record your own performance and deliver it in a different voice that you own or have licensed.
  • Your own cloned voice. Train a model on your voice so you can produce narration faster.
Tip
The honest money angle: businesses pay for outcomes, not tools. "I'll give your brand one consistent presenter across 30 videos" is a service. "I have FaceFusion installed" is not.
Checkpoint
You can name one use for each tool that you would actually do this month.
Stuck? Ask the community
STEP 03 of 06~2 min

How the three fit together

See the workflow that turns three separate tools into one character you control.

  1. Each tool stands on its own. Put together, they give you a complete on-screen character without hiring anyone.

    • 1. The brain writes it. Use GPT4Free to draft the script, the hooks and the captions.
    • 2. You perform it. Film yourself reading the script. Your timing and expressions are what make it feel human.
    • 3. The face goes on. FaceFusion replaces your face with your character's face in the recording.
    • 4. The voice goes on. Run your audio through the voice changer so the character sounds the same every time.
  2. That is one person producing a presenter-led channel from a bedroom. The modules after this one set up each tool in that order.

Checkpoint
You can describe the four-stage flow: write, perform, face, voice.
Stuck? Ask the community
STEP 04 of 06~3 min

What your computer needs

Know before you start which tools your machine will run well, and which will be slow.

The biggest factor is your graphics card (GPU), the chip that does the heavy maths for AI. A recent NVIDIA card makes two of these tools far faster. Without one they still run, just slowly.

  • GPT4Free: any computer. The AI runs on someone else's servers, so even an old laptop is fine. You need Python and an internet connection.
  • FaceFusion: any computer, but a graphics card changes everything. We ran it on a laptop with no NVIDIA card: a 5 second clip (121 frames) took 12 minutes 27 seconds, about 6 seconds per frame. With a modern NVIDIA card the same job takes seconds. It needs Python 3.12, Git and FFmpeg, and a few gigabytes of disk space for its models.
  • Voice Changer: a graphics card matters most here, because it has to keep up with you live. The project ships separate builds for PCs with and without an NVIDIA card, and for Macs with Apple chips. Use headphones so the changed voice doesn't feed back into your microphone.
Tip
No graphics card? Start with GPT4Free today, run FaceFusion on short clips overnight, and treat the voice changer as the one to test last.
Checkpoint
You know which of the three your computer will run comfortably.
If it breaksHow do I check which graphics card I have?
How do I check which graphics card I have?
On Windows press Ctrl + Shift + Esc, open the Performance tab and click GPU. The name is in the top right. On a Mac, open the Apple menu, then About This Mac.
Still stuck? Ask the community
STEP 05 of 06~2 min

The rules that keep you out of trouble

Know exactly where the line is, because these tools make it easy to cross.

  1. The reel was honest about it: these tools are legal, and that's exactly why you need rules of your own. Four cover almost everything.

    • Only use a face or a voice you own or have written permission to use. Your own, an AI-generated one that isn't a real person, or one you've licensed.
    • Never pretend to be a real person. Using a fake face or voice to get money, access or trust from someone is fraud, whatever tool made it.
    • No sexual content of real people, ever. In the UK and many other countries, making or sharing it is a crime.
    • Say when it's AI if it matters. Platforms including Instagram, TikTok and YouTube have rules on labelling realistic altered media. An advert or a news-style video needs the label more than a comedy sketch does.
  2. For GPT4Free the rule is different: your messages travel through services you don't control. Never paste passwords, customer data or anything private into it. Using a service through a route its owner didn't intend can also break that service's terms, so keep it for learning and experiments, and pay for an official API once something earns money.

    Warning
    FaceFusion has a built-in filter that blocks explicit content, and its licence forbids harmful use. Don't look for ways around either.
Checkpoint
You can state the four rules, and you know what you must never paste into GPT4Free.
Stuck? Ask the community
STEP 06 of 06~1 min

Where to start

Pick your first install so you get a result today.

    • Module 2: GPT4Free. Start here. It works on any computer and you'll have a real AI answer on your screen in about ten minutes.
    • Module 3: FaceFusion. Next. Install it, run your first face swap and learn the settings that make it look real.
    • Module 4: Voice Changer. Last. Download the right build for your PC, get your first changed voice, use it on a call and learn how voices are trained.
  1. Each module opens with a free step, so you can see exactly what you're getting before you unlock the rest.

Checkpoint
You know which module you're opening next.
Stuck? Ask the community

Your first wins

It works. Now do something with it.

You know the real sources

Three official GitHub links, so you never install a fake copy.

You know what each tool is for

One clear job for the face, the voice and the brain, and how they chain together.

You know the line

Four rules that keep you legal, and one for keeping private data out of GPT4Free.

Undo it cleanly

How to remove everything this mission installed.

Nothing was installed in this module, so there's nothing to undo. Each setup module ends with its own removal steps.

Next mission · The Lethal StackThe Lethal Stack, Module 2: GPT4Free, AI models with no API key
→