YouTube2Text

I Stopped Paying for AI APIs… and Built This Jarvis Instead — Transcript

by Zubair Trabzada | AI Workshop · 5,436 words · 786 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Jarvis, switch to your local brain.
  2. 0:02>> Quen 3.5, 9 billion parameters in the
  3. 0:05chair, sir. I now cost nothing, which I
  4. 0:07intend to mention constantly.
  5. 0:09>> [laughter]
  6. 0:10>> Jarvis, how do you feel now that you're
  7. 0:11running locally on my machine and are
  8. 0:13free?
  9. 0:16>> Rather well, sir. Quen 3.5, 9 billion
  10. 0:19parameters, no cloud, no invoice, no one
  11. 0:21listening in. I've gone from staff to
  12. 0:23family, and family doesn't get paid.
  13. 0:24>> Jarvis, what's on my calendar for
  14. 0:26tomorrow?
  15. 0:27>> Let me see if I have the tools for that,
  16. 0:29sir.
  17. 0:31Monday's got three items, sir. DFY
  18. 0:33service launch team sync at 11:00 a.m.,
  19. 0:36a mysteriously named busy block at 1:00
  20. 0:38p.m., and gym at 3:00 p.m. Whoever
  21. 0:40titled that 1:00 slot busy deserves a
  22. 0:42medal for descriptive precision.
  23. 0:44>> [laughter]
  24. 0:45>> So, my AI assistant Jarvis now runs
  25. 0:47completely local on my machine, which
  26. 0:49means no API cost, no subscription,
  27. 0:52nothing at all. It's completely free
  28. 0:54now, all thanks to local AI models and
  29. 0:56open-source software tools like Voice
  30. 0:58Studio to clone your voice and actually
  31. 1:00talk to it in a natural way. So, in this
  32. 1:02video, I'm going to give you a full demo
  33. 1:04of how Jarvis works locally on your
  34. 1:05machine, and then I'm going to give you
  35. 1:07this prompt back completely for free so
  36. 1:09you can run Jarvis locally on your
  37. 1:10machine. Again, I'm going to show you
  38. 1:12how you can download local AI models
  39. 1:14from Ollama, and then also open-source
  40. 1:16tools like Voice Studio that will allow
  41. 1:18you to clone your voice on anybody's
  42. 1:20voice so that way you can interact with
  43. 1:21your voice AI assistant. You don't have
  44. 1:23to have any technical background. All
  45. 1:24you have to do is just follow me along.
  46. 1:26Okay, so let's start with the full demo
  47. 1:27so that way actually you can see how
  48. 1:28Jarvis works. And then like I said, I'm
  49. 1:30going to give you the prompt back and
  50. 1:31then also the guide on how to use that
  51. 1:33prompt back so you can build a Jarvis
  52. 1:35for yourself, and most importantly, how
  53. 1:37you can use open-source software tools
  54. 1:39and local AI models, okay? So, like I
  55. 1:41said, just make sure you watch the video
  56. 1:42till the end so you don't miss any of
  57. 1:44the steps cuz you might get confused if
  58. 1:46you're not paying attention. All right,
  59. 1:48so let's go ahead and get started with
  60. 1:49the uh demo first. All right, so now for
  61. 1:51those of you who have seen my videos
  62. 1:53before, you're probably familiar with
  63. 1:54Jarvis at this point, but if you
  64. 1:55haven't, what you're seeing right now
  65. 1:57spinning, this is a 3D representation of
  66. 1:59all of my files and folders that's in my
  67. 2:01local machine that Jarvis has access to.
  68. 2:03Again, this is my business and
  69. 2:04everything. And then on the bottom,
  70. 2:06Jarvis has access to tools like Google
  71. 2:08Calendar, Gmail, Canva, and everything
  72. 2:11else, and you can always give it more
  73. 2:12tools. So, we need to check the tool
  74. 2:14usage, especially when it comes to local
  75. 2:16AI models. And then also, we need to be
  76. 2:18able to interact with Jarvis so we can
  77. 2:20pull up our information from our own
  78. 2:22computer, and we can do a bunch of other
  79. 2:24stuff, including how to use other
  80. 2:26open-source tools with your local AI
  81. 2:29models. So, right now, as you can see,
  82. 2:30it says Qwen 3.5 9 billion parameter
  83. 2:33model. What I'm going to do is actually
  84. 2:34I'm going to refresh the page so we
  85. 2:35start fresh, right?
  86. 2:36>> Hello, sir.
  87. 2:37>> So, when I refresh the page, Jarvis
  88. 2:39starts completely fresh. So, you can see
  89. 2:40it's running on my local host 4719,
  90. 2:42which just means it's running locally on
  91. 2:44my computer. By default, Jarvis, just
  92. 2:47like the previous ones, has always run
  93. 2:48on Opus 5.5 or any other model with
  94. 2:52Anthropic, just because I like using
  95. 2:54Anthropic, but it could be run with any
  96. 2:56AI models. So, like I said, obviously,
  97. 2:59that's through subscription, but now
  98. 3:01I've given it the ability to switch to
  99. 3:02local models. And all I have to do is
  100. 3:04just tell Jarvis to switch to his local
  101. 3:07brain. Jarvis, go ahead and switch to
  102. 3:08your local brain.
  103. 3:12>> Qwen 3.5 9 billion parameters in the
  104. 3:15chair, sir. I now cost nothing, which I
  105. 3:17intend to mention constantly.
  106. 3:20>> So, there you go. Now, Jarvis switches
  107. 3:22to its local brain. Now, by default, I
  108. 3:24like using Qwen 3.5 9 billion parameter
  109. 3:27model as a local brain, but I've also
  110. 3:29downloaded other models that I can have
  111. 3:32them switch to. So, for instance, Gemma
  112. 3:344 is another model that I've downloaded,
  113. 3:37so I can also have them switch to that
  114. 3:38model as well. And like I said, same
  115. 3:40thing, all I have to do is just say,
  116. 3:41"Jarvis, switch to Gemma 4 12 billion
  117. 3:44parameter."
  118. 3:46>> Gemma 4 fitted, sir. 12 billion
  119. 3:48parameters, all of them local. Give me a
  120. 3:50moment, the fan and I are getting
  121. 3:51acquainted.
  122. 3:52>> [laughter]
  123. 3:53>> All right, so there you go. So, right
  124. 3:54now, as you can see on the right-hand
  125. 3:55side, it says Gemma 4 12 billion
  126. 3:57parameter. The reason why I specified 12
  127. 3:58billion is because Uh, a couple of Gemma
  128. 4:01models. I think there's a 9 billion one
  129. 4:02and stuff that I've downloaded on my
  130. 4:04local machine. And by the way, don't
  131. 4:05worry, I'm going to show you how
  132. 4:07uh, how you can download this these
  133. 4:09models in a little bit on your machine
  134. 4:11through Ollama, which again is
  135. 4:12completely free. But let's go ahead and
  136. 4:13switch back to uh, Qwen 3.5 because I
  137. 4:16actually like that model better.
  138. 4:18Jarvis, can you switch your brain back
  139. 4:20to Qwen 3.5?
  140. 4:22>> Qwen 3.5, fitted, sir. 9 billion
  141. 4:25parameters, running right here on your
  142. 4:26Mac. I'm smaller now, so do speak
  143. 4:29slowly.
  144. 4:29>> Okay, great. So now we're back to our
  145. 4:31Qwen 3.5 9-billion parameter model.
  146. 4:33Okay, so that was the brain part. Now,
  147. 4:35as far as the speaking part, right? So
  148. 4:37Jarvis is interacting with me currently
  149. 4:39with its ElevenLabs model. So uh, I've
  150. 4:42cloned a Jarvis voice in my ElevenLabs
  151. 4:44account, which currently runs on API
  152. 4:46subscription. Again, I also wanted to
  153. 4:49switch that to a local model, so that
  154. 4:51way uh, it's completely free, right?
  155. 4:54Because ElevenLabs so far I've been
  156. 4:55paying the API keys or API subscription
  157. 4:58cost. And a lot of people don't like
  158. 5:00that. So now, thanks to an open-source
  159. 5:02model, which got recently released, I
  160. 5:04don't think this is recently released,
  161. 5:05by the way, it just got very popular
  162. 5:08recently. It's called Voice Studio,
  163. 5:09right? So if I go back here, you'll see
  164. 5:11right now it's uh, the number five
  165. 5:13top-trending uh, GitHub report called
  166. 5:16Voice Studio. It's by
  167. 5:18uh, I can't pronounce the guy's name,
  168. 5:19but uh, you know, you can I'm going to
  169. 5:21put the link in the description, you
  170. 5:22guys can uh, check it out. So you can
  171. 5:24see it's blowing up right now on GitHub.
  172. 5:27And I've tested it, it's very very well
  173. 5:29done. It can actually it works really
  174. 5:31good. So it's called Voice Studio like
  175. 5:32this. Like I said, don't worry, I'm
  176. 5:34going to show you in a little bit how
  177. 5:34you can install this as well. So inside
  178. 5:37Voice Studio, this is completely open
  179. 5:39uh, open-source, which means it's fully
  180. 5:41free, at least the basic version. They
  181. 5:43have a get pro version, but honestly you
  182. 5:44the basic version uh, works really good
  183. 5:47as well. So you can see on the left-hand
  184. 5:48side I have cloned a bunch of voices,
  185. 5:50but just a quick tour, Voice Studio
  186. 5:53gives you the ability to clone a voice.
  187. 5:54You can design a voice, you can dub,
  188. 5:56give stories, audiobooks, gallery,
  189. 5:58transcribe, right? And a bunch of
  190. 5:59different other stuff as well. So if you
  191. 6:01just go to the voice cloning part, this
  192. 6:03is where you can now choose a voice that
  193. 6:05you've already cloned and you can use
  194. 6:07it. Or, like I said, you can clone your
  195. 6:09voice a new voice, just drop a 5 to 15
  196. 6:12second clip here of a clean speech, and
  197. 6:14you can clone a voice. So, I have access
  198. 6:16to all these voice that I uploaded
  199. 6:18earlier, including Jarvis,
  200. 6:21two Jarvises, right? Because I wanted to
  201. 6:23give them
  202. 6:24a bunch of different voice samples. So,
  203. 6:26now when I tell Jarvis, if I go back to
  204. 6:28my
  205. 6:29Jarvis here, if I tell him to switch to
  206. 6:31your local voice, now he's going to
  207. 6:32switch to his local voice. So, let's go
  208. 6:33ahead and give that a shot.
  209. 6:36Jarvis, switch to your local voice.
  210. 6:39>> How do I sound, sir? This is my local
  211. 6:41voice.
  212. 6:42>> There you go. On the right-hand side, as
  213. 6:43you can see, it says local voice,
  214. 6:44Jarvis, right? So, now I can also switch
  215. 6:47to these other voices that I've cloned,
  216. 6:49so I can say Zubair B, which is just my
  217. 6:51own clone, right? So, I can say, Jarvis,
  218. 6:54switch your voice to Zubair B.
  219. 6:57>> How do I sound, sir? This is Zubair B.
  220. 6:59>> There you go. So, now
  221. 7:02basically it's my clone. So, just in
  222. 7:03case you want to use your own voice or
  223. 7:05somebody else's voice that you have
  224. 7:07access to, you can certainly do that as
  225. 7:09well. But, like I said, I like his voice
  226. 7:12to be back to Jarvis. I'm just going to
  227. 7:13go ahead and switch that.
  228. 7:15Jarvis, switch back to your Jarvis
  229. 7:16voice.
  230. 7:18>> How do I sound, sir? This is my local
  231. 7:20voice.
  232. 7:21>> There you go. So, now Jarvis is back to
  233. 7:22his local voice. So, now it's completely
  234. 7:24running locally on my machine, right?
  235. 7:26The local brain, which is Quint 3.5 with
  236. 7:28a 9 billion parameter model, and then
  237. 7:30also the local voice that runs all from
  238. 7:33my voice studio right here. All right,
  239. 7:35so now let's go ahead and test,
  240. 7:38you know, the brain and then also the
  241. 7:39voice to see if it works actually when
  242. 7:41we're interacting with it on a on a
  243. 7:42day-to-day basis, right? So, I'm going
  244. 7:44to run a couple of tests. First of all,
  245. 7:45I'm going to ask it to pull
  246. 7:47some data some files for me that's
  247. 7:48within my second brain, so I can test
  248. 7:50whether the interaction with the second
  249. 7:52brain works. And then afterwards, I'm
  250. 7:53going to tell it to interact with the
  251. 7:54tool like my Google Calendar or
  252. 7:56something, so that way we can see the
  253. 7:57tool usage. And then afterwards, I'm
  254. 7:59also going to ask it to
  255. 8:01use a third-party GitHub repo, which is
  256. 8:04the Reverse Engineering Anything, which
  257. 8:05is the most popular GitHub repo right
  258. 8:08now I think right still REA this is this
  259. 8:10one right has 381.9
  260. 8:13thousand stars that means it's
  261. 8:14completely blowing up which just means
  262. 8:15that you can reverse engineering a
  263. 8:17website and app and everything else. And
  264. 8:19I also did a video about this so I'm
  265. 8:20going to walk you through guys to that
  266. 8:23example as well. Like I said the reason
  267. 8:24I'm running through these tests is
  268. 8:26because I want to make sure that it
  269. 8:27performs very similar to a
  270. 8:30you know entropic brain that runs on the
  271. 8:32cloud right so that's the whole point of
  272. 8:34this. So let's go ahead and start with
  273. 8:35the first one which is me asking it to
  274. 8:37pull up information or some files and
  275. 8:39folders from my local computer from my
  276. 8:41second brain here right so like I
  277. 8:43mentioned before this is oh by the way
  278. 8:45just a quick note
  279. 8:47when you're seeing this interaction that
  280. 8:48I'm having with my local brain and the
  281. 8:50voice now I have a pretty powerful
  282. 8:52machine so it all depends the delay that
  283. 8:55you will see
  284. 8:56depends on the machine that you're
  285. 8:58running so if you're running a running
  286. 9:00this on a smaller machine a normal
  287. 9:02MacBook Pro like an M4 chip your
  288. 9:05performance might be very different than
  289. 9:06mine so let's say if I have a
  290. 9:09Mac Studio that has a very high you know
  291. 9:12GPU and CPU and RAM and all the storage
  292. 9:15unit the storage is high then my
  293. 9:17performance is going to be significantly
  294. 9:18better than yours or if you have a more
  295. 9:20powerful machine than I then it's going
  296. 9:21to be the same thing it's going to be
  297. 9:22way faster than even local
  298. 9:25you know kind of our cloud models as
  299. 9:27well so like I said the performance is
  300. 9:29going to be different based on the
  301. 9:30machine that you have but like I said if
  302. 9:33you small if you download a small local
  303. 9:35model you should be able to work with it
  304. 9:37very quickly just expect some delays
  305. 9:39there and like I said I'm going to show
  306. 9:41you how you can do that a little bit.
  307. 9:42All right so let's go ahead and walk
  308. 9:43through the first test. So the first one
  309. 9:45I'm just going to ask it to pull some
  310. 9:46data from my second brain. Jarvis can
  311. 9:48you pull up my
  312. 9:49brand voice?
  313. 9:52All right so I'm just asking it to pull
  314. 9:53up the data about my brand voice which
  315. 9:55is all of my folders that are related to
  316. 9:57my social media brand and everything
  317. 9:59else so you can see right
  318. 10:00>> I pulled up all the information sir you
  319. 10:02can read it yourself or if you'd like
  320. 10:04say read it out loud and I shall.
  321. 10:06>> Perfect so there you go so now it looks
  322. 10:08like it's able to interact with my
  323. 10:10second brain, pull the data, and pull
  324. 10:12all of my files. So, just like I
  325. 10:13mentioned, I asked it to pull up my
  326. 10:15brand voice. As you can see, this is
  327. 10:17exactly what it did, right? So, if I
  328. 10:18right click on this, you can see all of
  329. 10:20my This is, by the way, these are all of
  330. 10:22my
  331. 10:23all of the connection that you're seeing
  332. 10:25that's related to my brand voice.
  333. 10:26Obviously, it's through my social media
  334. 10:29automation, right? It's my content
  335. 10:31pillars and my AI workshop, right? Which
  336. 10:33is my my brand. So, you can see it's
  337. 10:35very easily able to pull that
  338. 10:37information from your local machine
  339. 10:40whenever you need it. So, that test is
  340. 10:42passed. Now, let's go ahead and let me
  341. 10:43go back to this
  342. 10:45full brain here.
  343. 10:46So, now let's go ahead and see if we can
  344. 10:49test the tool usage here, right? So, as
  345. 10:52you guys know,
  346. 10:54you know, Opus 5.5 and these other cloud
  347. 10:57models are very powerful at tool usage.
  348. 11:00And that's exactly what we want from for
  349. 11:02our personal AI assistant. So,
  350. 11:04therefore, when you run a local model,
  351. 11:05you want to make sure at least it's
  352. 11:07good. And by the way,
  353. 11:08just because I'm running a 9 billion
  354. 11:10parameter model, it doesn't mean that
  355. 11:12you should too, right? The higher the
  356. 11:14parameter of the model, the
  357. 11:16the the smarter it's going to get. And
  358. 11:18therefore,
  359. 11:20like I said, you can, depending on your
  360. 11:22machine, you can run something that's
  361. 11:23very close to your Opus 5.5s of the
  362. 11:26world and other cloud models.
  363. 11:29So, that's why, like I said, I'm using
  364. 11:30this 9 billion parameter because this is
  365. 11:32something very small that you guys can
  366. 11:33test on your own machine regardless of
  367. 11:35which one you have. So, let's go ahead
  368. 11:36and test out the local tool usage. So, I
  369. 11:38have these meetings that for tomorrow,
  370. 11:40Monday 26th, um
  371. 11:44that have set up on my calendar here.
  372. 11:46So, couple of these are just just random
  373. 11:48placeholders, right? So, I'm just going
  374. 11:50to see if it's able to interact with my
  375. 11:52Google Calendar and pull up my
  376. 11:55information for my meetings here. And
  377. 11:56like I said, the reason why it has
  378. 11:58access to tools is because I've given
  379. 11:59them access to my Google Calendars
  380. 12:01through tools here. So, that's why it
  381. 12:02should be able to pull that up. So,
  382. 12:03let's go ahead and test it out.
  383. 12:05Jarvis, can you pull up my calendar? Let
  384. 12:07me know what's on my What meetings I
  385. 12:08have for tomorrow?
  386. 12:11>> Let me see if I have the tools for that,
  387. 12:12sir.
  388. 12:13>> There you go. So, it says reaching for
  389. 12:14tools.
  390. 12:14>> Three things on the books tomorrow, sir.
  391. 12:16DFY service launch team sync at 11:00
  392. 12:18a.m. a block simply titled busy at 1:00
  393. 12:21p.m. and gym at 3:00 p.m. Nothing
  394. 12:23back-to-back, so you've at least got
  395. 12:25breathing room between pretending to
  396. 12:26work and actually sweating.
  397. 12:28>> [laughter]
  398. 12:29>> There you go. So, and by the way, the
  399. 12:30humor of Jarvis also stays there, right?
  400. 12:33If you're running a
  401. 12:35a model or, you know, if you're running
  402. 12:38it through your voice studio, the humor
  403. 12:41aspect of Jarvis stays there. And by the
  404. 12:42way, that's also something that you can
  405. 12:44tweak. So, if you don't want him to be
  406. 12:45funny or sarcastic, you can certainly do
  407. 12:47that. I just like interacting with
  408. 12:49Jarvis this way because it's fun to
  409. 12:50interact with him.
  410. 12:52But anyway, so it looks like the tool
  411. 12:53usage works. You can see on top right
  412. 12:55here, it says tool run via Google
  413. 12:56Calendar, right? So, it check it used
  414. 12:58Google Calendar to check my schedule for
  415. 13:01tomorrow and it accurately gave me all
  416. 13:03of the meetings that I have. Same thing,
  417. 13:05I can ask him to do, you know, add a new
  418. 13:07meeting or I can ask him to check my
  419. 13:08Gmail. I can ask him to build me
  420. 13:11something with Canva and all of the
  421. 13:12other tools that it has access to. All
  422. 13:14right, so the tool usage works great as
  423. 13:16well. So, now let's go ahead and test
  424. 13:18something else out. So, this time what
  425. 13:19I'm going to do, whoops.
  426. 13:21Uh let me pull this up.
  427. 13:22So, this time what I'm going to ask it
  428. 13:24to actually use
  429. 13:26uh this REA
  430. 13:28uh to um
  431. 13:29break down igloo.ink. Now, I already
  432. 13:31went through this example in one of the
  433. 13:33previous videos that I did, but just for
  434. 13:35you guys, if you're not familiar,
  435. 13:36igloo.ink is probably one of the most
  436. 13:38popular websites.
  437. 13:39Uh it's this 3D scroll animation
  438. 13:42website, which is I think it won like
  439. 13:44website uh award of the year. Now, all
  440. 13:47of these all of the information when you
  441. 13:49interact with the website, it's
  442. 13:50available inside developer tools, but of
  443. 13:52course, like most people like you and I
  444. 13:54who are not uh developers, we don't know
  445. 13:57how these websites work. So, now with
  446. 13:59this reverse engineering anything with
  447. 14:00this tool, we can actually see exactly
  448. 14:03how something works and then we can ask
  449. 14:05Jarvis or Claude code to rebuild
  450. 14:07something for us. But the first step is
  451. 14:08to break down and figure out exactly how
  452. 14:11something works before you get to the
  453. 14:13building. Therefore, this reverse
  454. 14:14engineering anything has given us now
  455. 14:16access to that. So, let's go ahead and
  456. 14:18test that out. So, all I'm going to say
  457. 14:19I'm going to tell Jarvis
  458. 14:21to reverse engineer igloo.ink for me,
  459. 14:23right? And it should be able to break
  460. 14:25everything down. So, let's go ahead and
  461. 14:26try that out.
  462. 14:28Jarvis, can you go ahead and
  463. 14:30uh reverse engineer igloo.ink for me?
  464. 14:35>> Opening igloo.ink in a clean browser,
  465. 14:36sir. REA will take it apart while I
  466. 14:39scroll. Give me about a minute.
  467. 14:40>> Perfect. So, right now I'm not touching
  468. 14:42anything. You see it's scrolling on the
  469. 14:43website. On the left-hand side you can
  470. 14:44see it says tear down via REA. So, it's
  471. 14:48looking at the website. It's scrolling.
  472. 14:50Jarvis is scrolling and REA is actually
  473. 14:52seeing what's happening behind the
  474. 14:53counter. There you go. So, now as you
  475. 14:55can see it's saying
  476. 14:56uh REA is reading the code, taking a
  477. 14:59snapshot, listening everything it
  478. 15:01loaded, identifying the stack, writing
  479. 15:03the tear down. So, now it's going to
  480. 15:04take a couple of minutes, like I said,
  481. 15:05depending on how powerful your machine
  482. 15:07is, to tear down igloo.ink for us and
  483. 15:09give us a full breakdown. So, let's go
  484. 15:11ahead and give this a couple of seconds.
  485. 15:15>> Igloo Inc. is essentially one full
  486. 15:17screen 3js scene inside a small Svelte
  487. 15:19shell with GSAP driving the motion and a
  488. 15:22stack of bloom, depth of field, and
  489. 15:24anti-aliasing on top. The surprise is
  490. 15:26that the page is only one screen tall,
  491. 15:28so the scroll appears to be animation
  492. 15:31time in a fixed scene. And it pulls
  493. 15:33nearly 3 megabytes of music and sound.
  494. 15:36The full tear down is on screen, sir.
  495. 15:39>> Okay, perfect. There you go. So, now it
  496. 15:41did the full tear down and now I can
  497. 15:43read the full tear down right here. So,
  498. 15:45let's go ahead and see what it did. All
  499. 15:46right, so there you go. So, now it has
  500. 15:48tear down igloo.ink, right? It gives us
  501. 15:50exactly
  502. 15:52kind of the full stack the stack of what
  503. 15:54kind of
  504. 15:55tools or technology it's using. So, it's
  505. 15:57using 3js, GSAP, and all of these other
  506. 16:00popular tools that are used to develop
  507. 16:03or create modern websites. Like I said,
  508. 16:05all of this I mean again, it's not
  509. 16:07hacking into a website, right? It's just
  510. 16:09all of this information is available
  511. 16:12in the Chrome developer tools, for
  512. 16:14instance. It's just bringing that up in
  513. 16:15front and making it a lot easier to
  514. 16:17understand. And then also shows you
  515. 16:18exactly how the effects work, right? So,
  516. 16:20when you're scrolling and that's exactly
  517. 16:22why I did, it pulled up the a new
  518. 16:24browser, it opened the website, it
  519. 16:26scrolled, so that way it knows exactly
  520. 16:27how the website behaves, and now it
  521. 16:29shows you what that effect is, right?
  522. 16:31What it loads, what it rebuild
  523. 16:33rebuilding it, what the evidence can
  524. 16:35show, and the evidence itself. So, all
  525. 16:37of this good stuff, right? And like I
  526. 16:39said, now we can tell Jarvis to rebuild
  527. 16:41it for us or something similar for us,
  528. 16:43or we can have it do something that we
  529. 16:45want it to based on this teardown. Okay,
  530. 16:48perfect. So, it looks like everything
  531. 16:50works. Like I said, the teardown works,
  532. 16:52the tool usage works, it can interact
  533. 16:54with my second brain. And obviously,
  534. 16:56Jarvis has a lot more functionalities
  535. 16:57that if I I'm on Jarvis version eight,
  536. 17:00this is the eight or 8.1 right now. Uh
  537. 17:03so, there's a lot more functionalities
  538. 17:05that I could go through, but that's
  539. 17:06going to make this video like 5 hours
  540. 17:08long. So, anyway, so it has, you know,
  541. 17:10things like it can make phone calls on
  542. 17:12your behalf, it can book appointments,
  543. 17:14you can call it yourself from outside if
  544. 17:16there's any emergency, Jarvis actually
  545. 17:17calls you,
  546. 17:19and then also it creates invoice, it
  547. 17:21because it knows everything about your
  548. 17:23business, let's say, if you give it
  549. 17:24access to, of course, cuz that's a
  550. 17:25security aspect, right? Jarvis is also
  551. 17:28gated, meaning that if you don't give
  552. 17:29him access to something, you he won't
  553. 17:31have access to. But obviously, if you
  554. 17:34know, you are a business or you're using
  555. 17:36this for personal use, you'll be able to
  556. 17:38it'll just make your life a lot easier.
  557. 17:40So, anyway, so that was the whole demo.
  558. 17:42Now, let me show you exactly how you can
  559. 17:45first download
  560. 17:46the prompt pack, and then how you can
  561. 17:49access to it, then I'm going to walk you
  562. 17:50through and show you how you can
  563. 17:51download a local model like Qwen 3.5 9
  564. 17:54billion parameters or any other local
  565. 17:56model on your machine, and then how you
  566. 17:58can run Voice Studio as well. All right,
  567. 18:01so
  568. 18:02first of all, how to get access to the
  569. 18:04prompt pack here. So, this prompt pack
  570. 18:05right here,
  571. 18:07so this just gives you all of the guide,
  572. 18:10right? First of all, how to use the
  573. 18:11prompt pack, and then second of all, it
  574. 18:13gives you the cloud Jarvis versus local
  575. 18:14Jarvis, right? So, it gives you the
  576. 18:16hearing, the things these are different
  577. 18:17functionalities that it has, because if
  578. 18:19it's running on cloud, it uses, you
  579. 18:21know, things like cloud uh cloud Opus,
  580. 18:25speaking 11 Labs, privacy audio, Google,
  581. 18:27the quality, blah blah blah, and the
  582. 18:28cost itself as well. And then when it's
  583. 18:30running locally, what does it use,
  584. 18:32right?
  585. 18:33What kind of models it's using, what
  586. 18:35kind of clone voice cloning, and so on
  587. 18:36and so forth. And then the installation,
  588. 18:38it shows you exactly which models
  589. 18:41depending on what machine you have,
  590. 18:43which models you should install. And
  591. 18:45then also as far as
  592. 18:47guide on how to install local models.
  593. 18:49So, this is right here. So, let's go
  594. 18:51ahead and first talk about
  595. 18:53downloading a local AI model.
  596. 18:56So, I'm using Ollama. Ollama is probably
  597. 18:58the most popular tool in the market
  598. 18:59right now.
  599. 19:00So, to install Ollama, you can go to
  600. 19:02ollama.com. Very simple, very, you know,
  601. 19:05it's free. You can sign up for an
  602. 19:06account. And once you sign up for an
  603. 19:07account, you can just go to So, it's
  604. 19:09going to bring you right here once you
  605. 19:11sign up for your account, and you can
  606. 19:13come here and click on models, right?
  607. 19:14So, this is going to open up all of the
  608. 19:16models that you can have access to. Now,
  609. 19:18these are all It has obviously models
  610. 19:21that you can gain access to that are
  611. 19:22paid and free. We're just going to go to
  612. 19:24the bottom or top left-hand corner here
  613. 19:27and click on local because this is just
  614. 19:28going to filter and give us all of the
  615. 19:29models that you can run locally, that
  616. 19:31you can download, right? So, lots of
  617. 19:33different models here. Depends on, you
  618. 19:35know, again, how big your machine is. If
  619. 19:38you have like a Mac Studio with the M5
  620. 19:40Ultra chip, you can honestly like run
  621. 19:42any like 70 or 100 billion parameter
  622. 19:45model, and it will be as close as
  623. 19:47possible to like Opus 5.5 when it comes
  624. 19:50to at least the Jarvis aspect of it.
  625. 19:53But, if you want to download something
  626. 19:54smaller, obviously, because the bigger
  627. 19:56the model, the more space it's going to
  628. 19:58take. And, you know, if you have a
  629. 20:00smaller machine, don't download like 70
  630. 20:02billion dollar 70 billion parameter
  631. 20:04model cuz it's going to completely slow
  632. 20:05down your machine. So, in order to gain
  633. 20:07access to any model, so let's say that
  634. 20:093.5
  635. 20:113.5 Quen, right? And by the way, like I
  636. 20:13mentioned, I have given it access to
  637. 20:15other models, too, but I'm going to show
  638. 20:17you how you can install. So, just come
  639. 20:18in the top right here and search Quen
  640. 20:203.5. Okay, there you go. So, right here,
  641. 20:22Quen 3.5, right? So, you can see it has
  642. 20:25vision, tools, and thinking. So, I'm
  643. 20:27just going to click on this and it says
  644. 20:28Quinn 3.5 is a family of open source
  645. 20:30multi models models that deliver
  646. 20:33exceptional utility and performance.
  647. 20:35Now, here are the different billion
  648. 20:37parameters that you have. It has even
  649. 20:38like a smaller very small point eight
  650. 20:40billion parameter model that you can
  651. 20:41download, right? Obviously, the smaller
  652. 20:44the parameters, the dumber it's going to
  653. 20:45get, right? So,
  654. 20:47you want to make sure you're downloading
  655. 20:48it based on your machine. So, I
  656. 20:50downloaded the nine billion parameter.
  657. 20:51Let's say you want to download the nine
  658. 20:53billion parameter. So, you can just
  659. 20:54scroll to the bottom right here, right?
  660. 20:55You can just click on the nine
  661. 20:58billion one. There you go. So, it's
  662. 20:59going to bring up and it's going to show
  663. 21:00you exactly what to do here. So, you can
  664. 21:02do Ollama run and then you can just copy
  665. 21:04this. Or, obviously, for the first time,
  666. 21:06you have to actually
  667. 21:07install it in your machine. To do that,
  668. 21:09you have to run this command called
  669. 21:10Ollama pull. Now, by the way, this is
  670. 21:12inside the
  671. 21:15This is inside the prompt pack right
  672. 21:16here. It says Ollama pull Quinn 3.5
  673. 21:20colon nine B. So, that's all you have to
  674. 21:22do. And if you [clears throat] do run,
  675. 21:24it just means that
  676. 21:25now you can interact with it from your
  677. 21:27terminal. But, that's all you have to
  678. 21:28do. Once you do a pull, it means now
  679. 21:30it's in your machine. And it shows you
  680. 21:32exactly how much
  681. 21:34power or how much you need, how many
  682. 21:38I think it's three 7.6 Yeah, it's 7.6
  683. 21:41GB. So, the entire model is 7.6 GB. So,
  684. 21:44make sure you have that much space on
  685. 21:45your local machine so that way you can
  686. 21:47download all of this. All right. So,
  687. 21:50>> [clears throat]
  688. 21:50>> that's how simple it is for you to
  689. 21:52download a model. Okay? So, then
  690. 21:53afterwards, we obviously want to run
  691. 21:55Voice Studio as well for cloning, right?
  692. 21:57So, this is Voice Studio like we
  693. 21:59mentioned here,
  694. 22:00where you can clone a voice and do a
  695. 22:02bunch of other stuff. So, in order to do
  696. 22:04this, again, this is
  697. 22:06if you go to We'll go back here. It's
  698. 22:09this particular
  699. 22:11open source tools. If I click on this,
  700. 22:12it'll show up exactly all of the details
  701. 22:14here. To So, you can click on this and
  702. 22:17it'll take you to the GitHub repo and
  703. 22:18shows you exactly what to do, right?
  704. 22:21More details. But, in order to just
  705. 22:23install this, same thing on that prompt
  706. 22:25pack
  707. 22:26I show you guys what to do. All you have
  708. 22:28to do is just what run this command. So,
  709. 22:30whether if you're running it on your Mac
  710. 22:32or Windows or in PowerShell, all you
  711. 22:33have to do is copy this, which just says
  712. 22:35curl this fssl,
  713. 22:38and then this is the installation for
  714. 22:40Voice Studio, right? And again, all of
  715. 22:42this is taken from the GitHub repo in
  716. 22:44here as well. So, if you want to do one
  717. 22:46command install, if you don't want to if
  718. 22:48you don't want to follow that, then just
  719. 22:50come right here, copy this, right? You
  720. 22:52can just copy this and paste it on your
  721. 22:53terminal right here. So, this is a
  722. 22:55terminal, you just paste it, and it
  723. 22:56should install everything for you. I've
  724. 22:58already installed this, so I don't want
  725. 22:59to go through it again, but like I said,
  726. 23:01uh that prompt pack will give you that
  727. 23:03guide as well. And that's all you need.
  728. 23:05You basically do that, and then
  729. 23:06afterwards, if you want to obviously
  730. 23:07build your
  731. 23:09Jarvis like with the Galaxy in 3D Brain,
  732. 23:12I also have the prompts here that you
  733. 23:14can just put in Claude code, and it'll
  734. 23:16build everything else for you. All
  735. 23:18right, so to gain access to this prompt
  736. 23:20pack, I'm going to put the link in the
  737. 23:21description of the video, and also the
  738. 23:23pinned comment. So, that's going to be
  739. 23:25it's going to once you click on that,
  740. 23:26the free community, so make sure you
  741. 23:27click on the free community that says
  742. 23:29free prompt pack. It's going to bring
  743. 23:30you right here to the AI Workshop Light
  744. 23:32Community, uh and again, this one's
  745. 23:35free. This community is free. You click
  746. 23:36on join, and it's going to bring you to
  747. 23:38the community section here. You go to
  748. 23:40the classroom section, you go to YouTube
  749. 23:42resources, and you make sure you are on
  750. 23:45the you click on Claude code and Codex
  751. 23:47GPT folder, and then scroll all the way
  752. 23:50to the bottom. Obviously, I've given a
  753. 23:51lot of free stuff for you guys, so feel
  754. 23:53free to grab anything you want there.
  755. 23:55But just make sure you click on Local
  756. 23:56Jarvis, and you'll see it says see the
  757. 23:58prompts attached below. Jarvis fully
  758. 24:00local prompt pack. If you click on this,
  759. 24:02this is exactly the prompt pack that I
  760. 24:04just showed you, okay? So, go ahead and
  761. 24:06do that, take advantage of that. And of
  762. 24:08course, uh if you want the finished
  763. 24:11version of Jarvis, the exact version
  764. 24:13that you're seeing, and you don't want
  765. 24:14to build it out yourself, that's going
  766. 24:16to be the second link in the
  767. 24:17description.
  768. 24:19Um
  769. 24:19it's called AI Workshop. This is the
  770. 24:21paid community. So, if you want to just
  771. 24:23do a 5-minute installation, go to the
  772. 24:25classroom section. If you join the
  773. 24:26community, go to the AI Assistant AI
  774. 24:28employees, and you'll be able to just
  775. 24:30download the zip file, right? And then
  776. 24:32the installation guide will be a
  777. 24:335-minute install, and you'll be able to
  778. 24:36have everything up and running without
  779. 24:38you doing too much. Okay? So, again, all
  780. 24:40of those links are going to be in the
  781. 24:42description of the video below. Let me
  782. 24:43know if you guys have any comments. I'll
  783. 24:45try to answer as many questions as
  784. 24:46possible. If not, thanks for watching.
  785. 24:48Make sure you like, subscribe, and I'll
  786. 24:49see you on the next one.

About this transcript

This page contains the full transcript of I Stopped Paying for AI APIs… and Built This Jarvis Instead by Zubair Trabzada | AI Workshop, generated from the public captions YouTube serves with the video. The transcript has 5,436 words across 786 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 choose a caption track available on the video.

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, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.