Big Idea

The argument in one line.

A free open model can produce a finished faceless video if a well-built skill carries the knowledge, because the skill, not the model's intelligence, is what makes the result good.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A beginner who wants to try faceless short-form videos without paying for a voice tool, an animation tool or an API key.
A creator who has seen faceless channels pull tens of millions of views and wants to test the format before spending anything.
Someone who has not used a coding agent like Claude Code or OpenCode and needs the setup explained step by step.

SKIP IF…

You need to produce videos at channel scale every day, since the free usage limit and the 30-second length make that unreliable.
You only have a phone, because everything runs on a computer with a project folder.

TL;DR

The full version, fast.

Faceless channels earn millions of views, and tools for them charge up to $149 a month. The creator builds one for $0 using a free open model, GLM 5.3 Flash, inside OpenCode. The workflow has three parts: write a roughly 60-word what-if story in free ChatGPT to save free usage, install OpenCode and pick the free model, then load a skill that handles the look, characters, voice, sound and self-checking of frames locally. Direct it like an editor with timestamped notes and watch the output yourself. The catch is an unpublished free usage limit; paid OpenCode Go is about 60 cents a video.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 01:19

01 · The hook and the $0 claim

Proof of faceless channel views, a sample 28-second video, and the promise of no subscription, credits or API key.

01:19 – 02:37

02 · Roadmap and community winners

Outline of the steps, the skill as the key piece, and a showcase of community posts.

02:37 – 04:15

03 · Step 1: write the story in ChatGPT

Why the story is written outside the free model, and what-if prompts to try.

04:15 – 04:44

04 · Step 2: set up OpenCode

Install, pick GLM 5.3 Flash free, set the effort level.

04:44 – 05:39

05 · Step 3: how to direct it

First version, notes to the agent, and four tips.

05:39 – 07:04

06 · Step 4: the skill

Why the skill matters, what is inside, and how to install it and run /new.

07:04 – 08:18

07 · Step 5: the catch

Free usage limits, OpenCode Go pricing, and a note on using ChatGPT to save tokens.

08:18 – 09:41

08 · Viewer questions and close

School access, Higgsfield, Windows, phones, hardware and other agents.

Atomic Insights

Lines worth screenshotting.

The model is not the magic in an AI video workflow; a skill that already knows the look, voice and checks lets a weaker free model deliver.
Write the story in a separate free chat tool first so the free coding model spends its limited usage only on building the video.
Start with 30-second videos and run one at a time; running two at once hit the free usage limit in under an hour.
Give notes to an AI agent the way you would to an editor, with a timestamp, such as 'at second five, change this'.
An agent saying everything looks good is not proof; watching the render yourself still catches errors it missed.
A faceless video can run entirely on a laptop: voice, drawing, sound effects and render are local, and only the model is remote.
Free open models are free for a limited time and their usage caps are undocumented, so treat them as a way to test, not a channel foundation.
OpenCode Go at $10 a month works out to about 60 cents per video at roughly 100 videos.
A what-if premise with short, one-idea sentences is the story shape that works for a 30-second stick-figure video.

Takeaway

A good skill beats a smarter model

WHAT TO LEARN

A free model can finish a video when a skill holds the style, voice and checks, so the work is splitting jobs so the free brain only builds.

Write the story in a separate free chat tool so the coding model's limited free usage goes to the build, not the writing.
A skill packages the look, character behavior, voice and self-checks, which lets a less capable model produce consistent results.
Direct the agent like an editor: short notes with timestamps, such as 'at second five, change this', beat vague feedback.
Start with 30 seconds and one video at a time, since running two at once triggered the usage cap in under an hour.
Watch the render yourself, because the agent can report that everything looks good while the video is still wrong.
Free open models have undocumented, shifting limits, so use them to test an idea and budget for a paid plan if it works.

Glossary

Terms worth knowing.

OpenCode: A free, open-source coding agent that runs in your terminal, reads your files and runs tools on your computer, similar to Claude Code or Codex.
GLM 5.3 Flash: A free open model available inside OpenCode, used here as the brain that writes and runs the video build.
Skill: A folder of instructions and tools that teaches an AI agent how to do one job, such as making a video in a set style.
Faceless video: A video with no on-camera presenter, built from animation, voice-over and sound instead.
WSL: Windows Subsystem for Linux, which OpenCode recommends on Windows for running its terminal setup.

Resources

Things they pointed at.

01:19toolOpenCode

01:19toolGLM 5.3 Flash

02:37toolChatGPT (free tier)

06:14productFaceless Videos Skill v2 (5 styles)

07:33productOpenCode Go

08:50toolHiggsfield

Quotables

Lines you could clip.

05:49

“Everyone thinks the model is the magic. It's not.”

Clean contrarian claim with no setup needed→ TikTok hook↗ Tweet quote

06:03

“The skill already knows how to make the video, and it makes the model check its own frames before you ever see them.”

Explains why a weak model can deliver→ newsletter pull-quote↗ Tweet quote

00:31

“What if you got paid for every word you said?”

A ready-made story hook→ IG reel cold open↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

look at this video 30 million views and this one 15 million same channel rick's clips and it's not a one -off their shorts have over 300 million views you guys asked for this i went through the comments on our videos and you kept asking me the same thing can i make faceless videos without a cloud subscription can i do this for free so i tried it and it looks like this tomorrow every word you say pays a dollar good morning two dollars You say about 16 ,000 words a day, almost 6 million a year.

You quit your job. Then one morning, alone, the counter stops. Only pays if someone is listening.

That night, your partner's on their phone. I love you. The counter doesn't move.

Would you still say it? That's a faceless video, 28 seconds with the voice, the animation, the sound, horizontal and vertical. And it was built by your free open model GLM 5 .3 flash inside open code.

Like I did for my AI editor last time, this one is for faceless. There are tools that charge up to 149 a month for this, and you could do it for $0. No subscription, no credits, no API key, no 11 labs.

You don't have to create an account. So today I'm going to show you how to make yours step by step. step.

I know a lot of you are not technical, so I'm not skipping anything. First, we have to write a story in ChatGPT or cloud, and I'll show you why that saves you free usage. Then we set up open code.

Then we'll show you how to direct it. And at the end, I'll give you the skill. That's the most important part of this video.

Because without it, you're not going to get the results. And there's also one catch about open models. And I'll tell you all about it.

And I read all your comments. So I'm answering those at the end too. If you're new here, my name is Jad.

I give away skills for free three to four times a week in my school community. Link is in description. We do weekly calls.

to help you with the skills. So if you get stuck on anything today, just join the call. And this video is not sponsored by open code or GLM.

Nobody paid me for this. I just read the comments and you guys are asking for this. And this week I picked the posts in school where you guys took my skills and shared what you made.

Thank you for doing that. It really means a lot. I want to show them here and I'll put everyone's tag on screen.

So hopefully this can support your content too. So here are this week's winning posts. In July of 2026.

A photograph appears on the internet. One glass of wine wrecks you for two days, but your friend has three and feels fine. You're home alone and it's 11 at night.

Someone knocks on your front door. So without further ado, let's get straight into it. So real quick here, the first full version took me about 34 minutes.

And that's with me changing my mind halfway there and making it short. And to be fully blunt, OpenCode's own stats say the total cost is $0. And here is why it's free.

The brain is the only thing that isn't on my computer. And the brain is free right now. The voice, the drawing, the sound effects, the render, all of that runs on my MacBook.

So you need three things. ChatGPT for the story. Open code for the build and the skill, which I'm saving for last.

Step one, try to be T -first. Here's why. Everything the free model writes uses your free usage.

And you want to save that for the part that actually builds the video. Writing the story is super simple. You can either do it yourself or you can give it to ChanGBT.

You only need to use the free version of ChanGBT and it will work completely fine. So you asked it for 30 seconds. What if story for a stick figure video about 60 words, short sentences, and one idea each.

I put the full prompt in the description. So just copy it. Here are some things that you can try.

What if you got paid for every word you said? That's the one you just watched. What if a zombie apocalypse started today?

What if everyone knew the day they would die? Or tell Chajibiti your niche. Ask for 10 what ifs and pick the ones you would click on or scroll through your feed in the middle of the night.

Then keep the text. We're going to need it at the end. Step two, open code.

It's like cloud code or codex. It reads your files, runs tools on your computer and keeps working by itself. It's free and open source.

Copy the line from my description in school. You paste it in your terminal. You hit enter.

On the windows, open code recommends WSL. I will also give a link in the school commune. day in the learning section.

You open a new terminal, you type open code, hit enter, no sign in. If you see connect, just skip it. Then you type slash models, search free, pick GLM 5 .3, flash free.

Press control and T until you see high. And that's it. You now have a free AI agent on your computer.

But right now it has no idea how to make a video. that's what the skill is for and we're getting there first let me show you how to direct it because that's what makes the video good this is the first version i got before any notes for a first try at zero dollars and this is genuinely crazy now the title is hard to read and this money counter is the whole story so it should be way bigger so it got notes the same way you would give notes to an editor make the counter bigger 11 notes and about 11 minutes later i got the results for Now, let me give you the tips that I've learned the hard way.

The first tip, start always with 30 seconds and make one video at a time. If you try to do two at a time, you will hit the wall. Tip two, give it your full story and your shots, not just one lap.

That's what I tested and that's why we did ChatGPT first, if you remember. Tip three, give notes with timestamps. For example, at second five, change this.

Tip four, watch the video yourself. Sometimes it told me. Everything looked good, but when I watched it with my own eyes, it was still wrong.

Okay, so now the skill, the part that I told you about. Everyone thinks the Marvel is the magic. it's not a free model isn't as smart as claude but here's the thing it doesn't have to be because the skill already knows how to make the video it's a folder with everything i learned making these the look of the world how the characters act the voice the sound how to write for the ear and it makes the model check its own frames before you ever see them that's why free model can do this And the voice runs on your computer.

So there's no voice subscription and no API key. And as you know me, I'm not gatekeeping any of this. It's completely free.

There are five styles in there right now that you can install. The one I tested on the free model is the stickman. So that's where I would genuinely start.

Go to my school, open the post for this video, open the Google drive and download the faceless video zip. Leave it there. Don't unzip it.

go back to open code and paste the install prompt from the drive. It says cloud code inside. Don't worry.

Open code reads the same folder. It sets up the voice, the tools about one gigabytes, wait for it self -test past, and you can skip the music tomorrow. That one is 15 gigabytes.

Now type slash new, paste your story and your shots from chat UBT and say, make this a 30 second stick man video, horizontal and vertical. It shows you a plan and you just click on go. And there's your first.

faceless video for $0. All right, now the catch that I told you about. Open models are free for a limited time, and there is a free usage limit.

I couldn't find their number anywhere in their documents, so you can't use it for a long period of time. And I wouldn't run it fully on a channel. And it actually happened to me.

I had two videos running at once, and that's honestly on me. And in under one hour, I got a free usage limit reached. Prime.

What's strange, last week, I was able to run it for two hours with no problem. So it looks like OpenCode keeps changing.

Because if you want to do it, you have to do it now because it's still free till this day. And when you outgrow him, it's cheap. Open code go is $10 a month.

And by my math, a hundred videos like this, or you pay as you go, that would cost you 60 cents per video. I also want to be transparent. You remember how I said, use ChediBT to write a story.

I did the same thing for my cloud to remind me of the story, to save tokens on the open code, to make this video. So I completely understand why you want to use an open model. When you're starting, you don't know if you want to start.

And I think this is the great way to start to see how crazy ai can be for a free model completely free but if you have a budget i would really recommend paying for a subscription now your comments i read them and thank you so much for commenting that's why the videos exists so let me answer what you guys are asking My school says pending.

Do I have to pay? The school is completely free. We accept people by hand and there's over 12 ,000 of us now.

I'm sorry. It takes me a little bit of time to accept everyone. I just have to do it manually.

Do you need Hicksfield? So no, you don't really need Hicksfield because I've already made style sheets. I have like, I think around five, but if you want to have your own style sheets, you can use their API.

You pay as you go. So you pay for one image. And the next question, does it work for windows?

Yes. it also works in chrome if it's installed can you do this on your phone i'm sorry you can't everything runs on your computer it needs the files it needs a folder it cannot do it on the phone so what computer do you need you don't really need a strong one as long as you have space because the brain is the thing that is really heavy and we're not installing no brain we're just using open code what about codecs or anti -gravity i haven't tested those yet i'm not going to promise you anything um but it is on to -do list So that'll be it.

And you can comment on this video too. Just so you know, I read them and that's where my next video actually comes from. Tell me what you would like to see next.

And I would love to see you in the community and join the call. Link is in the description. Make video, post it.

I'll try to push your video too, as you guys are helping me. If you're not subscribed, please leave a follow. I drop skills three to four times a week.

Thank you for sticking all the way to the end. And I'll see you in the next one.

The Hook

The bait, then the rug-pull.

Faceless channels pull tens of millions of views, and the tools that make them cost up to $149 a month. Jad shows a $0 version built on a free open model.

Frameworks