用开源模型 GLM 5.3 Flash 和自建 Skill 零成本剪辑视频的完整教程

以下是视频口播的自动转写，可能有错字。

This video cost me zero dollars to edit. No editor, no claude, no ChatGPT. A free open model did all of it. And look, i love opus 5.05.

I made so many videos about it. But sometimes those credits, they keep adding up. ChatGPT has the same story too. And a real editor is not cheap.

So i wanted to know one thing. What happens when you give a free model the right guidance? Because i have never seen a single youtube video.

Of a free model. So i did it. No massive prompt, no prompting for hours, just my skill. A tiny prompt in my video. That video literally costed me zero dollars.

And when i say zero dollars, i mean it literally zero. No claude subscription, no ChatGPT, no credits, no api key. I didn't even make an account or had to put my card somewhere.

I used a free open model called glm 5.3 flash. So what i did before doing this video, i went online and tried to see if somebody has ever done it before.

Sure, i have seen people using open models. For code, but not for ai editing. I think i might be literally the first person on youtube doing this.

So, in this video, i'll show you why this wasn't possible before, and why it is now. Which of my editing skills work with a free model, and how to install it properly step by step.

I know a lot of you are not technical, some are. So, i'm not skipping anything. Time stamps are below too, if you want to jump right in.

If you're new here, my name is jad. I give away skills for free 3~4 times a week, in my free school community. Link is in description.

And we do have weekly calls with my business partner, where we help you with the skills or answer any questions that you have.

So, if you get stuck on anything today, just join the call, we would love to have you. And also want to mention this video is not sponsored by open code or gllm.

Nobody paid me for this. I just saw the school community, a lot of you were talking about open models. So let's get straight into it.

So first, why does this really matter? Because i have been there. When you just starting, money is a bottleneck. Cloud is 20 dollars a month, and on top of that, you have credits.

Every test you run, every edit, every mistake, it costs you something. And when you start, you don't even know if this is going to work out for you.

So you're paying just to learn. I know exactly how that feels, and having a real editor that's even more expensive. So a lot of people just don't start, they wait until they can afford it.

So now with the help with ai, there are no more excuses for us. I want to explain why this wasn't possible a week ago. Before, if you wanted an ai to edit your video, you needed an agent, something that can open your files, run tools on your computer, and keep working for an hour without you.

Something like Claude Code, and to use Claude Code, you need at least 20 dollar plan. And yes, there were free models, but they were either small enough to run on your laptop or or not small.

Smart enough to do a full edits or they just weren't good enough to keep working that long without getting lost. What changed is three things all coming together.

The first one is open code. It's like Claude Code and ai agent that lives in your terminal but it's free and open source.

And here's the part that you might miss open code reads the same skill folders as Claude Code. So my Claude Code skills that have worked on and that have spent a lot of times.

I didn't have to rebuild it. The second thing, open models became free inside it on october 16th, open code made gllm 5.3 flash free.

And this is not a tiny mole. It's a 321 billion parameter model. It's open mit license. Anyone can download it. It got downloaded more than 6 million times on hugging face and the last month alone, the best part, even if you have a bad computer, you don't have to run it locally yourself.

The third thing, the skill and this is the real secret. My skill editor v1 holds my whole editing system and this is where i saw a bit of confusion in my last video.

People thought i had a master prompt and i had to go back and forth. That's the old way. Nobody does that anymore. Now we build skills to be able to just give a small tiny prompt or not a prompt at all.

You just give it the video and it figures out itself. So it's holding the same editor style that i used for cloud editor v1.

It sees how often the picture changes the colors, the sounds where the caption sits and the rules. It's not allowed to break.

So i don't have to prompt it. So the model doesn't have to be genius, it just has to follow the recipe or need guardrails.

Again, that's why you don't need a massive prance. Whoever tells you, oh, you need to have a massive prong to be able to have a faceless content or to edit your videos, that's complete bs.

And if you put those three all together, and the free model can actually edit a video, as like you just saw. Now, i want to be super transparent with you.

Everything in this edit ran on my macbook. The cutting, the transitions, the graphics, the rendering, the sound mixing, all in my laptop.

But the brain, the model itself runs on open code servers for free, because 321 billion parameters does not fit my computer.

I'm very far from that, but this is genuinely what gets me excited. These models are open, and open models keep getting better and smarter every month.

I already run a small open model on this exact laptop right here, 1 3.5. It's 9 billion parameters. I use it for my car sales, either when i take a flight, and i don't have wi-fi, i can use my ai.

I haven't tried it for editing, because honestly, it's too small to edits. But it's generally a matter of time before it can do these things for me.

All right, so this is how it works. If you're new here, i have already given editing systems, my screen share editor, my short editor, and the new style editor v1 and the motion graphics.

The one that you just saw, i built all of them with cloud, but they also work in open code, as i mentioned earlier. Because the hard part where i strongly recommend not doing is building the skill in open code.

Cloud is much more smarter, it's it's better to build the skill in cloud, and that's what i did. I built in cloud, and then gave it to you, so you can use an open code, and it wouldn't cost you anything.

So once the skill is built, using it is a lot easier, and the model just follows it. So, if you have all my editing styles that i gave away, it works in almost all of them, except my motion graphic skill.

The one where every frame is made by code and the music, that's really like heavy code, and there is no open model that is able to do that right now.

So, may, when i made that video, all i did is type edit this video for me, and i dragged my video, i put it in the chat, and i said use the editor v1.

It used 1.6 million tokens and 0 dollars it costed. Is it as good as opus 5.5? No, but it definitely does the work. So now, how do you install it properly step by step?

So just to give you a quick heads up, it works on mac, windows, open code recommends wsl, which is basically linux inside windows.

I'll put the link in description. Step one, install open code on mac. Press command and space, type terminal, and open it.

Then copy the install line from my description. By the way, the description will be in my school. It's the one from open code dot ai.

Paste it in the terminal and hit enter, and use the exact line from my description, because that's the version i tested.

The second step, in my school community, you'll see that i've already shared an editor v1. You go and open the google drive, and download the editor v1 zip.

Leave it in your download folder. Don't move it, you don't even. Even need to unzip it, the setup will find it by itself.

Step three, close your terminal, open a new one, and type this time, open code, and hit enter. You don't need to sign in, if you see something called connects, just skip it.

That's for paid models. Step four, type slash models, search free, and pick g l m 5.3 flash. Free, then press control and t, that changes how hard it fixes.

Keep pressing until you see max. That's what i used, because editing is a big job. Step five, open the setup prompt from the google drive, copy everything, paste it into open code, and hit enter.

It says Claude Code inside it, don't worry about it. Open code reads the same folder, it finds the zip, installs the skill, and then installs all the free tools for you.

Ffmpeg, python, the video render, and the speech model that listens to your voice. If something is missing, it tells you, and it asks you before anything that needs your password.

It also asks you one question. Do you want the cut out effects? That's where you stand in front of the background with a huge.

Word behind your head. That one is about 1 gigabyte. So if you have space to say yes, then it will make a 7 second test video to prove everything works.

If you see it, you're done. Step 6 type slash new start a fresh session. So the skill loads again. Step 7 drag in a video that you want to edit and the prompts that you have to put in edit this video with editor v1.

You click on enter. That's literally my prompt that i used and i think you can use it too. When it says it's done, just reply to it, open it for me and watch it and give it notes.

The more you give it notes, the better it gets. And that's it, that's zero dollars right there. Now, here are my tip, the first tip that i really recommend, start short.

My video was 38 seconds. Don't drop a 20 minute video on your first try. Tip two, when you give notes, use timestamps. Say like at second 5, not these second 5 points.

Some reasons it did it didn't do that for me well, and tip three, before you send the prompt to edit your video, ask it how long will it take?

It tells you before it starts. And that's it, a full edit with animations, real sources and sounds made by a free open model that is zero dollars, and it's only going to get better from here.

So, if money was the thing that was a little bit blocking you, i think this is the best way to start content. The skill is free in my school community, link is in description, install it, edit it, and i would love to see your post in school.

I genuinely want to see what you make, and if you get stuck, just join the calls or leave a post, now we'll reply to it, as i am active in the school community.

Tell me in the comments what would you like to see a free open model do, or a skill that you want, and i can create it and then.

Give it for free. If you're not subscribed, please leave a follow. I drop skills every three to four times a week. Thank you for sticking all the way to the end.

It really means a lot to me. And now, see you in the next one.