I Tested PewDiePie's FREE ChatGPT - Odysseus AI (Setup + Guide) — Transcript
Full transcript
- 0:00PewDiePie just dropped his free open
- 0:02source AI workspace for your computer
- 0:04and it looks and it feels like Claude or
- 0:07even ChatGPT. We'll pair it with Ollama
- 0:10that runs the AI model on your hardware.
- 0:12Odysseus runs everything that you see
- 0:15[music] and touch. Your chats, your
- 0:16research, your documents, your notes.
- 0:18You can load almost any open model out
- 0:21there. For anyone using the basics of
- 0:23ChatGPT or Claude, you can now do that
- 0:26for free. And the best part about all of
- 0:28this is that you are not renting it.
- 0:30There's no subscription involved. You
- 0:32are not signing into anybody's cloud and
- 0:35the things that you actually type in, it
- 0:37stays on your machine. And in today's
- 0:40video, we'll set up Odysseus AI. We are
- 0:42going to test its features and we are
- 0:43going to see if it's actually worth
- 0:45time. And super quick, if you get stuck
- 0:47anywhere, then our free to join school
- 0:49community has every command and
- 0:51troubleshooting steps right there
- 0:53waiting for you. The link is down in the
- 0:55description below. Go and check it out.
- 0:57And now,
- 0:58let's check out Odysseus. Now, the first
- 1:01thing that we need to do is we need to
- 1:02open up our terminal and I know that
- 1:04word makes some people so nervous. So,
- 1:07let me just kill that right now.
- 1:08Terminal is just a text box where you
- 1:11type in instructions to your Mac or your
- 1:13PC and that's basically it. You are
- 1:16talking to your computer in a slightly
- 1:19different way, okay? So, the first thing
- 1:22that we need to do is we need to install
- 1:23Homebrew. And if you've never heard of
- 1:26it, just think of it as the App Store
- 1:29but for command line tools. It makes
- 1:32installing everything else is so much
- 1:34more simple. Now, head over to brew.sh
- 1:37and then grab the install command. And
- 1:39then you're going to paste it into
- 1:40terminal and then you're going to hit
- 1:41return and it'll take just like a couple
- 1:43of minutes and then it's going to ask
- 1:45for your password along the way. And
- 1:47now, once Homebrew is done, all you have
- 1:49to do is just run the command on the
- 1:51screen and then it installs Git, Python,
- 1:53and Tmux all in just one shot. Now, just
- 1:55a quick note, I'm not going to be saying
- 1:57all of the commands out loud in the
- 1:59video because you actually don't need to
- 2:01type all of this in from scratch. All of
- 2:03the commands are down in the Free School
- 2:05community. All you have to do is just go
- 2:06over there, copy and paste them, and
- 2:08then you can just follow along with the
- 2:10video, okay? Remember, link is down in
- 2:12the description. Now, quick breakdown so
- 2:14that you know what you've installed. Git
- 2:17is what we'll use to download Odysseus.
- 2:20Python is what Odysseus actually runs
- 2:23on, and then tmux is basically a small
- 2:26helper that Odysseus uses for some of
- 2:29the background tasks. Now, when it
- 2:31finishes, you are good, okay? Just keep
- 2:33going, and now we're going to download
- 2:35Odysseus itself, and the way we do that
- 2:38is called cloning, which I know sounds
- 2:41pretty technical, but it just means that
- 2:42we are copying the entire project from
- 2:44the internet onto your computer. Then
- 2:47you're going to run the two lines that's
- 2:48on screen, and the first one copies the
- 2:50whole project onto your machine, and the
- 2:52second moves you into that new folder.
- 2:54And that part is extremely important
- 2:56because every command after this runs
- 2:59from inside it. Now, just take a second
- 3:01to let that sink in. PewDiePie's code,
- 3:03the interface, all of it, is now sitting
- 3:06on your machine, and it was that easy to
- 3:08set everything up. Okay, so now I want
- 3:11you to stop because before you launch
- 3:14anything, there's one thing that you
- 3:16have to do first. And almost every
- 3:18tutorial about this has skipped it so
- 3:20far, which I mean, to me, is exactly why
- 3:22people end up in the comments just
- 3:25swearing the app is broken, and some
- 3:27people even saying that it just hangs
- 3:28forever and it never loads. Okay, so
- 3:31here's what's actually happening. The
- 3:33very first time Odysseus starts, it
- 3:35needs to download something called an
- 3:37embedding model. It's a small file that
- 3:40helps the AI understand and search text.
- 3:43Now, if your internet hiccups even
- 3:45slightly during the first launch, the
- 3:47download isn't going to stall, [music]
- 3:48and then the whole app just freezes
- 3:51mid-startup. Then your browser shows
- 3:53nothing and you just think that you
- 3:55broke something. When really it's just
- 3:56waiting on one file. Now the fix is
- 3:58simple. All we have to do is download
- 4:00the file ourselves before we launch the
- 4:02app. The commands are on screen right
- 4:04now. So just run the first three to set
- 4:06up the environment and then the last one
- 4:08to grab and cache the model. And that
- 4:11last one's basically a mouthful. So do
- 4:14yourself a favor, just go over to the
- 4:15free school community and then go and
- 4:17copy it over there and then you just
- 4:19paste it in, okay? It's going to save
- 4:20you a bunch of time when it comes to
- 4:21typing. Now you'll see a progress bar
- 4:23and then it prints cached, okay? And if
- 4:26your connection drops partly through,
- 4:29then it's pretty much a no big deal.
- 4:31Just press the up arrow and then hit
- 4:33return and then it picks up right where
- 4:35it left off. Now that one step just
- 4:38saved you a ton of frustration. So just
- 4:40going to say it right now, you're
- 4:41welcome. Now let's launch it, okay? So
- 4:45you're going to run the start script
- 4:46that's on screen right now and then
- 4:48you'll see the terminal start moving.
- 4:50Now the lines of text scrolling by,
- 4:52there's checks passing, the database is
- 4:54setting up. Then it will stop and it
- 4:55will ask you to create your admin
- 4:57account. It asks for username. So just
- 5:00type one or hit return to keep the
- 5:03default. And then it's going to ask for
- 5:05a password and then, okay, now another
- 5:08one, a heads up.
- 5:10When you type the password, nothing
- 5:12shows on screen. There's no dots, no
- 5:14stars and that is pretty normal. That's
- 5:16how terminal hides it. Just type it in
- 5:19and then hit return and then write that
- 5:22username and that password down to make
- 5:24sure that you always have it because you
- 5:26are going to need it every time you open
- 5:28the workspace. Now it prints setup
- 5:30complete and then it starts the server.
- 5:32Now because we cached that model in step
- 5:34three, it boots in a few seconds and
- 5:36then it opens your browser automatically
- 5:38to localhost colon [music] 7860.
- 5:42Then you're going to log in with the
- 5:44account that you just made and this
- 5:47this is now the moment, okay?
- 5:49This is your AI workspace. It's a
- 5:51running on your PC and it's not a server
- 5:55in another country, it's not a
- 5:56subscription, it's now yours. And I
- 5:59mean, if you just look at that for a
- 6:00second, you can see the navigation down
- 6:02the left, the chat window in the middle,
- 6:04and the model picker sitting at the
- 6:06bottom right. That's empty right now,
- 6:08but we are about to fix that. So, right
- 6:11now the workspace is up, but there's
- 6:13nothing behind it yet. There's no model
- 6:16to actually [music] answer you. Now,
- 6:18that's what Ollama is for. Go to
- 6:21ollama.com and then download the
- 6:24installer and then you're going to run
- 6:25it. It takes about a minute to finish
- 6:27and then when it's done, you'll spot a
- 6:29small icon up in your menu bar at the
- 6:32top of the screen. And when the Ollama
- 6:35is there, then Ollama is up and now it's
- 6:38ready to load models. Now, we pull our
- 6:41first model, the small M-135,
- 6:46and I want you to start with the
- 6:48smallest ones out there.
- 6:50Now, this is just to prove that the
- 6:52whole system works before we upgrade.
- 6:55And then we are going to run the command
- 6:57on screen. Now, that's a tiny model.
- 6:59It's about 19 MB, so it's going to
- 7:01download pretty fast. And [music] when
- 7:03it finishes, you've got something for
- 7:04the workspace to talk to. Now, we
- 7:07introduce the two programs to each
- 7:09other. Ollama's running, Odysseus is
- 7:11running, but they don't know about each
- 7:14other yet. And here's how you connect
- 7:16them. It's not buried in some settings
- 7:18menu, so that's at least one thing you
- 7:20don't have to worry about. It happens
- 7:22right in the chat. All you do is you
- 7:24click the chat box and then you type the
- 7:26setup command on screen and then
- 7:28Odysseus shows you your options. Now,
- 7:30for a local model like ours, it tells
- 7:33you to paste your endpoint address
- 7:35straight into the chat. So, that's what
- 7:38we do. We drop in the address on screen
- 7:40and we're going to drop that into the
- 7:41chat and then we are going to hit enter,
- 7:44and that address just points to Ollama
- 7:46running on your computer. It looks like
- 7:49a web link, but it actually never leaves
- 7:51your machine. Now, here's the mistake
- 7:54that gets almost everyone the first
- 7:56time. If you paste that before you've
- 7:59pulled the model, Odysseus comes back
- 8:01with no Ollama models found, and people
- 8:05basically just panic after seeing that.
- 8:06They think something broke, but the fix
- 8:09is exactly what we already did in the
- 8:11last step. You have to pull a model
- 8:13first. Now, on our side, we did that, so
- 8:16we are fine. Just paste that address in,
- 8:18and this time it comes back with found
- 8:21one model, and then it drops your state
- 8:23into a fresh chat with it selected. Now,
- 8:26if you ever need to switch to it
- 8:27manually, then that select model
- 8:28drop-down at the bottom right is where
- 8:30every model that you've pulled shows up.
- 8:33Now, just type your first message, and
- 8:35then watch. And that reply, it didn't
- 8:38travel anywhere. There's no API call,
- 8:40there's no round trip to OpenAI or
- 8:43Anthropic. Your machine generated every
- 8:45word of it. Great idea, and it's
- 8:47offline. Now, that tiny model, small
- 8:50135, was just to prove that the pipe
- 8:52works. It's honestly too small to be
- 8:55useful day-to-day, so let's grab
- 8:57something with some material range to
- 8:59it. So, I personally go with Qwen 2.5 7B
- 9:02if your PC's got the memory for it. It's
- 9:06sharp, and it's released under the
- 9:08Apache 2.0 license, which means it's
- 9:11genuinely free to use for anything,
- 9:14including your work. Now, on a lighter
- 9:16machine, drop down to Qwen 2.5 1.5B. It
- 9:21has the same license, a smaller
- 9:22footprint, and still a big step up from
- 9:25where we started. And then basically
- 9:26just pull whichever one fits. The
- 9:28commands are on screen right now. And I
- 9:30mean, now, let me just answer the
- 9:31question that I know you must be asking.
- 9:34Is this actually free? Well, the answer
- 9:36to that is yes. Not a tile, not a fee
- 9:39for now before we drop a price on it.
- 9:41This is free. They are open weight
- 9:44models with open licenses, so you
- 9:46download them once and then you run them
- 9:48as much as you want. There's no per
- 9:50message cost, no monthly bill. The only
- 9:52actual real cost is that bit of extra
- 9:54power that your machine draws while it's
- 9:56thinking. Now, once it finishes
- 9:58downloading, just go back to Odysseus
- 10:00and then open that select model drop
- 10:02down at the bottom and then just pick
- 10:03the model that you just pulled. And
- 10:05that's a no re-setup, no
- 10:07re-configuration, just pick it and then
- 10:09just go.
- 10:11Now, let me show you what you just
- 10:13built. Because once you see it all
- 10:15together, you realize how wild it is
- 10:18that this thing basically costs no money
- 10:21whatsoever. Now, first, the chat. It
- 10:24works like you'd expect. You type and
- 10:26then it responds, you have a
- 10:27conversation, except this one is
- 10:29private. Then you just run it with your
- 10:31Wi-Fi off and it still answers. Nobody
- 10:34else is reading it, no one is training
- 10:36on it, no one is keeping a copy on their
- 10:38servers. The only place it's saved is on
- 10:41your machine. And then there's deep
- 10:43research. You simply just give it a
- 10:44topic and this is the one feature that
- 10:46does reach out to the web. It pulls in
- 10:48sources, it reads them and it writes you
- 10:50up a report with citations. Now, simply
- 10:52just type on something that you actually
- 10:54care about and then watch it work
- 10:56through the whole thing and hand you
- 10:57something that you'd have spent an
- 11:00entire afternoon on yourself. Then
- 11:03there's also documents. You open a doc
- 11:05right inside Odysseus and the AI helps
- 11:08you as you write. It's basically like
- 11:10having a writing partner sitting right
- 11:12beside you, except you are the one
- 11:15that's driving everything. And then
- 11:17there's memory. Now, this is the one
- 11:18that pays off over time because it holds
- 11:21onto context. It stops feeling generic.
- 11:25So, just tell it about a project once
- 11:27and it will still know that exact
- 11:29project next week. The more you work in
- 11:31it, the more it fits the way you
- 11:34actually think and operate. And here's
- 11:36the thing, apart from deep research and
- 11:38reaching out for sources, all of that
- 11:40ran right on your machine for free
- 11:43without handing your work to anyone.
- 11:47Now, that's what you built today. And
- 11:49then there's tasks. This is where it
- 11:51starts doing things without you. You can
- 11:52set up scheduled prompts, research
- 11:54answers, or actions, and it just does
- 11:57them on its own. Have it run a research
- 12:00report on a topic every week. You set a
- 12:02trigger, a schedule, an event, a
- 12:04webhook, and then you just pick what it
- 12:06should do, and then it just runs in the
- 12:08background. It's kind of like giving
- 12:09your AI a to-do list, and then having it
- 12:12actually will follow through. Okay, so
- 12:14let me knock out the three issues that
- 12:16most people hit, and this is going to
- 12:17help your build go more smooth.
- 12:20If the app won't load or it just hangs,
- 12:23it's usually that embedding model from
- 12:25step three that's getting interrupted.
- 12:27So, just rerun that cache command, and
- 12:30then run the start script again, and
- 12:32then you're done. Now, if you see the
- 12:34command not found when you try an Ollama
- 12:37command, then it almost always just
- 12:39means that Ollama isn't installed or
- 12:42isn't running. So, then something just
- 12:44check for Ollama in your menu bar. Now,
- 12:46no Ollama, no model, just open Ollama
- 12:48first. And if you get a port already in
- 12:53use error, then something else is
- 12:55sitting on the port Ollama wants. So,
- 12:58the command that's appearing on screen
- 12:59right now, that clears it. And then
- 13:01simply just start it again. Now, I've
- 13:03dropped all of this down in the free
- 13:05school community. Go and check it out,
- 13:07and everything that we covered in
- 13:09today's video, all of it is down there,
- 13:11okay? So, the link is down in the
- 13:13description below. Now, simply just take
- 13:16a step back, and then look at what
- 13:17you've got. The model, the interface,
- 13:20and your chat history. All of it lives
- 13:23on just one [music] machine. It's yours.
- 13:26There's no bill showing up every month.
- 13:28There's no company sitting in the middle
- 13:30and nobody setting the rules about what
- 13:33you can ask the thing that you are
- 13:34running locally. And this is just the
- 13:37starting line. From here, you can pull
- 13:39bigger models as you get comfortable.
- 13:41You can set it up to run tasks for you
- 13:45automatically. You can wire it into your
- 13:47own tools and files and even your
- 13:49workflows. And if you want the full
- 13:52command list from today plus the
- 13:54troubleshooting guide and you want a
- 13:56place to ask some important questions,
- 13:59then the free school community is where
- 14:00you are going to want to be. The link is
- 14:03down in the description below. Go and
- 14:04check it out. There's a bunch of people
- 14:06that is currently there all learning
- 14:08from each other and you could join them.
- 14:10Okay. So now, if you made it till the
- 14:12end of this video, I'm pretty sure that
- 14:13you had a lot of fun. So, give this
- 14:15video a like, subscribe to the channel,
- 14:17and comment down below. What do you
- 14:19think the future holds for Odysseus? Do
- 14:21you think it's going to change
- 14:23everything? Or do you think it's going
- 14:24to burn up pretty soon? Let me know down
- 14:26in the comments down below. And also,
- 14:28check out one of the videos popping up
- 14:29on the screen right now because I
- 14:30guarantee you you're going to learn
- 14:32something cool and I will catch you on
- 14:35the next one.
About this transcript
This page contains the full transcript of I Tested PewDiePie's FREE ChatGPT - Odysseus AI (Setup + Guide) by Zinho Automates, generated from the public captions YouTube serves with the video. The transcript has 2,818 words across 395 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.