10 Tool AI GRATIS Così Potenti da Rendere INUTILI Molti Abbonamenti — Transcript
Full transcript
- 0:00Right now, you're likely paying for a
- 0:02pile of AI subscriptions, and it’s
- 0:03not exactly pocket change. Anyone
- 0:05working in this field knows it’s
- 0:07super easy to spend anywhere from $ 20
- 0:09to $ 300 a month on various AI
- 0:11subscriptions. I spend about €
- 0:13400-500 a month just on subscriptions,
- 0:15and my agency spends much more than
- 0:17that. I've tried practically every paid
- 0:19tool out there, so I know exactly what
- 0:21you’re paying for. Now, I won’t
- 0:23tell you that I’ve canceled
- 0:25everything because that’s not true,
- 0:27but a large chunk of what I was paying
- 0:29for—images, voice, coding, and so on
- 0:31—I’ve cut because each of those
- 0:33things now has a free, open-source
- 0:34version that does the same job and can
- 0:36even run on your computer. So today,
- 0:38we’re going to look at 10 free tools
- 0:40that you absolutely need to know, and I
- 0:42assure you they are truly insane, so
- 0:44let’s not waste time and get started.
- 0:46Open Montage, a complete video from a
- 0:48single sentence. The first one is
- 0:50called Open Montage, and it’s the one
- 0:52that really raised my eyebrows. Look at
- 0:54this. This is a complete cinematic
- 0:56video about a snowfall in the city at
- 0:58night. It uses real footage and has a
- 1:01voiceover explaining why snow dampens
- 1:03sound. I made it by writing one
- 1:05sentence, and it did the rest. Open
- 1:07Montage defines itself as the first
- 1:09open-source, agentic system for video
- 1:11production. And the right word there is
- 1:13"production," because it doesn't just
- 1:14generate an isolated clip. It takes
- 1:16charge of the entire video from start
- 1:18to finish. It does the research for you
- 1:20, decides how to structure the script,
- 1:22finds the footage, edits it, adds music
- 1:24, and renders it. It climbed to the top
- 1:26spot on GitHub Trending and gathered
- 1:28tens of thousands of stars in just a
- 1:30few months, which tells us how many
- 1:32people are watching it. The way to get
- 1:33it running is what I'll be using for
- 1:35almost the entire list. Open your agent
- 1:37—I use Cloud Code—paste the GitHub
- 1:39link, and tell it to download and
- 1:41configure it, because that’s the
- 1:43unique thing about Open Montage. The
- 1:45agent isn't just an accessory; it’s
- 1:47the director. You give the brief, and
- 1:49it executes. By the way, the prompts
- 1:51are in the description so you can go
- 1:53back and use them later. In this case:
- 1:55make me a 40-second cinematic video
- 1:56about a snowfall in the city at night,
- 1:59using only real footage. As you can see
- 2:01, it started, went to search for clips
- 2:02online, and downloaded quite a few. It
- 2:04edited everything with fades, did the
- 2:06color grading, and it took a few
- 2:08minutes. The first time I used it, it
- 2:10took up to 10 minutes, but in the end,
- 2:12we had a finished video. Then I pushed
- 2:14it a bit further and told it to add a
- 2:16voiceover explaining why fresh snow
- 2:18makes everything seem quieter. And here
- 2:20is the interesting part, it wrote the
- 2:22voiceover script all by itself. It
- 2:24chose a documentary tone because it
- 2:26understood the video's context,
- 2:28generated a narrative, and mixed it in
- 2:30while lowering the city noise
- 2:32underneath. I didn't write a single
- 2:33word of that script. Here is what the
- 2:35final result actually looks like.
- 2:37Have you ever noticed how quiet
- 2:38everything becomes after a snowfall? It
- 2:40is not just a feeling.
- 2:42I will tell you one thing right away
- 2:43because you won't hear it many more
- 2:45times in this video. The software is
- 2:47free, but the voices and models you
- 2:48connect to it might have usage costs,
- 2:50so that is very important. So if you
- 2:52stick to real footage and free voices,
- 2:55you stay at zero, otherwise not. Number
- 2:57two, Qwen Image. That was a whole video
- 3:00from just one line. Now let's move on
- 3:02to something simpler, but which solves
- 3:04a problem you have surely encountered.
- 3:06If you wanted a single image with text
- 3:08on it and you wanted that text to be
- 3:10written well, because the output of all
- 3:12image generators with text is awful and
- 3:14we know it. Number two solves this
- 3:17problem, which is Qwen Image, Alibaba's
- 3:20image model. Look at this first one, it
- 3:22is an open notebook on a wooden table.
- 3:24And on the page, it is handwritten:
- 3:27Chapter 1. Every letter is in the right
- 3:29place. Normally, you ask an image
- 3:31generator for a word and it gives you
- 3:32back a scribble or something that
- 3:34clearly looks like it was made with an
- 3:36app. The third generation came out just
- 3:38as I am filming, and they built it
- 3:40around one single promise. Images you
- 3:42can use, not just look at. In practice,
- 3:45this means text that is readable even
- 3:47when small, labels, signs, captions,
- 3:49and very long prompts of up to a few
- 3:51thousand words to compose a dense image
- 3:54in one go. Clearly, I will use simple
- 3:56prompts here to show you the speed and
- 3:58quickness, but you can go and create
- 3:59some truly crazy prompts. In this case,
- 4:02I recommend JSON. In this case, I said
- 4:04a shop window in a neighborhood at
- 4:06night with warm light with the sign
- 4:07open until late, and as you can see,
- 4:09the result is amazing and above all,
- 4:11very fast. The text is clean, the thing
- 4:13is beautiful. Two clarifications
- 4:15because they matter. You can use the
- 4:17new first version for free from the
- 4:19Qwen chat with a daily generation limit
- 4:21, but you cannot download them. It is
- 4:23the only tool on this list that does
- 4:25not run on your own machine. The second
- 4:27one, previous versions are actually
- 4:29open with a free license even for
- 4:31commercial use, and those you can take
- 4:33home. If you need to produce graphics
- 4:36in a burst without depending on anyone,
- 4:38start there. Tool number three,
- 4:40Chatterbox, your voice cloned for free.
- 4:42Now, something a bit different. What if
- 4:44you took your voice and cloned it for
- 4:46free? Number three is indeed Chatterbox
- 4:48. First, listen. Hello, my name is
- 4:50Gabri Solution, how are you? That voice
- 4:53isn't me, it's a clone of my voice
- 4:54built for free with this tool, and it
- 4:56sounds a lot like how I really sound.
- 4:59Voice cloning works like this: you give
- 5:01the AI a sample of your voice, it
- 5:03learns how it sounds, then you write a
- 5:05sentence and it says it with that voice
- 5:07. Usually, this stuff is paid. Tools
- 5:10like Eleven Labs, which is crazy, make
- 5:12you pay for it, but Chatterbox does the
- 5:14same job and is from Assemble AI. It is
- 5:17open source with an MIT license and
- 5:19runs on your machine. It’s the
- 5:20perfect example of the trick from
- 5:22before, because I didn't install
- 5:24anything by hand. I know, when the
- 5:25installation starts you see all that
- 5:27text scrolling, it's a bit overwhelming
- 5:29, but you don't have to touch anything.
- 5:31You copy the Git link and tell your
- 5:33agent to install this and launch it for
- 5:34me. You choose the folder, check the
- 5:36space, it downloads what's needed, and
- 5:38opens the app. Inside, it's trivial.
- 5:41You upload short clips of your voice.
- 5:4310 clean seconds are enough, you choose
- 5:45the language, write the text, and it
- 5:47generates it. In a minute, the voice is
- 5:49ready. It's multilingual; Italian is
- 5:51there, and it keeps the same voice even
- 5:53when you change languages. You record
- 5:55in Italian and make it speak in English
- 5:57while still being you. One very
- 5:59important thing to know: every file it
- 6:01generates has an inaudible digital
- 6:03watermark inside that serves to
- 6:05identify that the audio is synthetic.
- 6:07This is a very good thing because at
- 6:09least we are bulletproof and safe.
- 6:11Think of what else people could do with
- 6:14it, and above all, you can use it for
- 6:16voiceovers, or maybe you recorded audio
- 6:18for a video that doesn't work well and
- 6:20you want to redub it; in short, the
- 6:22uses are endless. Nemotron 3 Ultra, the
- 6:25brain that builds the tool for you. We
- 6:27have covered video, images, and voice.
- 6:30The next one is a real brain, signed by
- 6:32NVIDIA; I know, it sounds crazy. And it
- 6:35builds a working tool for you from just
- 6:36a sentence. This is indeed Nemotron 3
- 6:39Ultra, and look at what it created for
- 6:41me. This is a customer acquisition cost
- 6:43calculator. You input how much you
- 6:45spend on advertising, how many leads
- 6:47you get, and how many you close, and it
- 6:50tells you your cost per lead, cost per
- 6:52customer, and what you’re actually
- 6:54making. And the part that sold me is
- 6:57that after writing it, it went back to
- 6:59check its own work and corrected itself
- 7:01. I still have to prove it to myself;
- 7:03this is insane. But let’s take a step
- 7:05back first. Nemotron-3 Ultra is the
- 7:07largest open model Nvidia has ever
- 7:09released. 550 billion parameters.
- 7:13Released in June, it wasn’t built for
- 7:16chatting, it was built to power agents,
- 7:18meaning very long tasks where it has to
- 7:20perform 20 steps in a row without
- 7:22losing the thread. And Nvidia has
- 7:24opened up practically everything: the
- 7:26weights, the recipe, the data. It
- 7:28speaks Italian. Now, the part that
- 7:31almost no one tells you: a model like
- 7:33this won't run on your laptop. That’s
- 7:35clear. To run it at home, you need
- 7:38graphics cards that cost as much as a
- 7:40car, but it’s open, which means
- 7:42anyone can host it. In fact, you can
- 7:44find it for free at € 0 per million
- 7:46tokens on the services we’ll see in a
- 7:49minute. So, I connected it to my agent.
- 7:51In simple terms, it means that on the
- 7:53outside I’m still using Claude as
- 7:55always, but the brain doing the
- 7:57reasoning now is Nemotron. Then I gave
- 7:59it a simple prompt: build me a customer
- 8:01acquisition cost calculator, a single
- 8:04HTML file I can open with a double
- 8:06click. I input advertising expenses,
- 8:08number of leads, and so on. No servers,
- 8:11no installations, just one file. It got
- 8:13to work and didn't just stop after
- 8:15writing the code. It went back to read
- 8:17its own output. It found an error and
- 8:19corrected it by itself without asking
- 8:21me anything. When it opened, I had a
- 8:23tool in front of me: drag the sliders,
- 8:25change numbers, see images move—
- 8:27basically everything from one line of
- 8:30instructions. Five: Omnirot, free fuel
- 8:32for your agent. There is one thing,
- 8:35though: running an agent all day
- 8:37consumes tokens, and tokens cost money.
- 8:40The next tool gives it free margin, and
- 8:42it's also how you access Nemotron
- 8:44without paying anything. Number 5 is
- 8:47Omnirot. If you use AI agents for work,
- 8:50this could quietly be the most useful
- 8:52one on the list. You already know the
- 8:54problem it solves. If you run out of
- 8:56credits in the middle of a job,
- 8:58everything stops. Or you have keys
- 9:00scattered across four different
- 9:02programs, each with its own limit, or
- 9:04at the end of the month you get a bill
- 9:05bigger than you expected. Omnirot acts
- 9:09as a single local gateway on your
- 9:11computer, giving you access to nearly
- 9:13300 different providers from there. As
- 9:18I'm filming this, 90 of them have a
- 9:21free tier on their site, offering about
- 9:231.6 billion free tokens per month.
- 9:27That’s their number, not independent
- 9:29data, but the mechanism is exactly that
- 9:31. When you send a request, it picks the
- 9:33best model available among the free
- 9:34ones, and if one runs out, it
- 9:36automatically switches to another. Your
- 9:38agent basically never stops. There is
- 9:40also something I didn't expect. It
- 9:42compresses requests before sending them
- 9:44by cutting out the fluff without
- 9:46touching the code, and for heavy
- 9:48sessions, it saves a massive chunk of
- 9:50tokens. It has an MIT license with over
- 9:5220,000 stars and works with Cloud Code,
- 9:54Cursor, Codex, and practically all the
- 9:56others. Setting it up takes 3 minutes;
- 9:58it doesn't run locally like everything
- 10:00else. You enter the dashboard and
- 10:02create a key. I know, "API key" sounds
- 10:04technical, but it’s just a password
- 10:06you copy once, name it, and that’s it
- 10:08—you paste it into the agent you’re
- 10:10using, and from then on, everything is
- 10:12routed to the cheapest or free model
- 10:14without you having to think about it
- 10:15again. Now your agent has a free brain
- 10:18and free fuel to keep going. Number 6,
- 10:21Anarlog, the private notebook for
- 10:23meetings. Now let’s change gears
- 10:25completely. The next one has nothing to
- 10:27do with code; it's for anyone who
- 10:28spends their days in one call after
- 10:30another, which happens to me. Number
- 10:32six is Anarlog. It’s the project that
- 10:34many of you know as HyperNote. It
- 10:35changed its name, but it’s the same
- 10:37thing. It’s a note-taker for meetings
- 10:39, and it does three things. It listens
- 10:41to the call, writes the full transcript
- 10:43, and finally gives you a clean summary
- 10:45, all on your PC. And let me explain
- 10:47why that "on your PC" part carries so
- 10:48much weight. If you've used Fireflies
- 10:50or Granola, for example, you know two
- 10:52things. One, they are paid services,
- 10:54and two, all the audio from your
- 10:55meetings ends up on a server that isn't
- 10:57yours. For a lot of people, that’s a
- 10:59problem, especially for companies.
- 11:01Think about cases where you might talk
- 11:02about margins, price lists, or clients;
- 11:04that stuff cannot get out. With Anarlog
- 11:06, it doesn't leave, it stays on your
- 11:08device, and it’s open source with an
- 11:11MIT license, so anyone can go ahead and
- 11:13verify how the data is handled. There's
- 11:16another detail I like: no bot enters
- 11:18the call to stare at everyone from a
- 11:20black box. It just listens from your
- 11:23machine and works with any platform
- 11:24because it doesn't need to integrate
- 11:26with any of them. The installation is
- 11:28the usual Cloud plus link and off you
- 11:30go. I know, it's truly crazy. Give it a
- 11:33try. Number 7, Invoke. You have an
- 11:36image studio on your PC. Back to
- 11:38creation, the next one makes images
- 11:40directly on your computer and, above
- 11:42all, fixes them. Number 7 is Invoke AI.
- 11:45Look here, this photo doesn't come from
- 11:47any paid site, it runs on my computer
- 11:49and didn't cost me a thing. Invoke is
- 11:53the cleanest model for generating
- 11:54images locally without having to study
- 11:57anything. If you've heard of Stable
- 11:59Diffusion, this is the version with a
- 12:01well-made interface that opens in your
- 12:03browser. You're not there staring at
- 12:05code, but the reason I keep it
- 12:07installed isn't for generating, it's
- 12:08for the canvas. You can select a piece
- 12:10of the image—a poorly rendered hand,
- 12:12crooked text, or the background—and
- 12:14have only that piece regenerated.
- 12:16It’s an incredible thing. The rest
- 12:18remains the same. In practice, the work
- 12:20you really need—the problem isn't the
- 12:22whole image, it's almost always a
- 12:24detail, and this way you can fix it. On
- 12:26the prompting side, as always, complex
- 12:28prompts in JSON or just plain spoken
- 12:30language. Photograph of a wooden desk
- 12:32from above, open notebook, cup of
- 12:34coffee, natural morning light. And then
- 12:36you can also use negative prompts. No
- 12:38text, no logos. A nice little detail:
- 12:40you can also tell it what you don't
- 12:42want in the image and it understands
- 12:43very well, so you can even combine both
- 12:45types of prompts. And now let's move on
- 12:47to number 8. Open Generative AI, an
- 12:49entire studio with over 200 models. I
- 12:52know. And if you wanted a complete
- 12:54studio—images, video, audio—all in
- 12:56one place, I'm telling you, it seems
- 12:58impossible, but it is possible and it's
- 13:00the next one. Our tool is Open
- 13:02Generative AI and it is literally a
- 13:04creative studio that stays right there
- 13:06in your home. Look, this dashboard has
- 13:08over 200 models inside, divided by
- 13:10department: an image studio, a video
- 13:13studio, one for lip-sync, one for
- 13:15cinematic shots with lens and aperture
- 13:17controls, and one for building
- 13:19workflows where one thing transitions
- 13:21into another. Think of all the big paid
- 13:24platforms you know for generating
- 13:26images and video. This is the
- 13:28open-source version of those, put
- 13:30together in a setup that runs in your
- 13:32home. The models you find inside are
- 13:35big names: Flux, Cling, Sora, Veo,
- 13:37Sivim; in short, heavy-duty stuff. You
- 13:40install it locally like everything else
- 13:41and it gives you an address that opens
- 13:43in your browser. Inside, you open the
- 13:45text-to-video generator, choose the
- 13:46model, write the prompt, set the
- 13:48quality, and hit play. But be careful
- 13:50here, the interface is rightly free and
- 13:52open source. What you pay for is the
- 13:55studio, the convenience of having 200
- 13:57models in one place instead of six
- 13:59different subscriptions, but the models
- 14:02aren't theirs, so they run through an
- 14:04external service with your key and
- 14:06heavy generations and all the
- 14:07generations to be done are paid per use
- 14:09as usual. You will read a lot about
- 14:12this tool being completely free, but be
- 14:15careful, the interface is free. If you
- 14:17then go to individual video generation
- 14:19models, clearly you have to pay for the
- 14:21use of credits there. It is truly
- 14:22incredible. It reminds me a lot of
- 14:24Xfield. Only Xfield contains just the
- 14:26video part. In here, there really is
- 14:28everything. Number 9, Open Cut. Number
- 14:319, Open Cut is an open-source
- 14:33alternative to CapCut. In my opinion,
- 14:35even the name is a bit reminiscent of
- 14:37it. You know the reason why it exists.
- 14:39CapCut was born free, conquered the
- 14:40market, then month after month, piece
- 14:42by piece, it began to become more and
- 14:44more of a subscription and practically
- 14:46everything is now paid. Open Cut is a
- 14:48classic track timeline, cuts and
- 14:50exports with an MIT license, which is
- 14:52the structural guarantee. If one day
- 14:54someone put a subscription on it, the
- 14:56community would duplicate it in three
- 14:58days, identical, but without a
- 15:00subscription. There is a detail that
- 15:02matters a lot for my work. It opens in
- 15:04the browser and the files don't go
- 15:05anywhere. The editing happens on your
- 15:07computer, this is a very nice thing,
- 15:09with your processor, a client's video,
- 15:11perhaps under NDA or maybe your own
- 15:13videos you don't want circulating
- 15:15around. In fact, there has been a whole
- 15:17debate here too with CapCut regarding
- 15:18the use of videos edited there, in
- 15:20short, the issue of privacy and so on.
- 15:22Here, however, everything stays local.
- 15:24A warning: it is a young project, so
- 15:26the core of the editing works, but a
- 15:28part of what you expect inside CapCut
- 15:30isn't there yet. For very simple things
- 15:32, it's really excellent. For still
- 15:34complex things, it will take time
- 15:35because it was just born. Very nice,
- 15:37though, because it has a plugin
- 15:38architecture and the plans also include
- 15:40control via agents. In my opinion,
- 15:42let's keep an eye on it in a few months
- 15:44because this one really rocks. And now
- 15:46let's go to the last one, but one of
- 15:48the most important, number 10. Number
- 15:5010, Raymotion, from code to rendered
- 15:53video. Number 10 is Raymotion and it
- 15:55closes the circle. It is one of the
- 15:57engines that Open Montage, number one
- 15:59on this list, uses under the hood to
- 16:01build videos. Watch this clip first. As
- 16:05you can see, it's a counter going from
- 16:070 to 100,000 with typography, text, the
- 16:09moving number, and much more. It's a
- 16:12real video, created entirely through
- 16:14code by an AI without opening After
- 16:16Effects or needing a motion designer.
- 16:19Remotion is a framework that turns code
- 16:21into real video files. That might sound
- 16:23boring, but wait, because it’s
- 16:25actually insane. It's made for the kind
- 16:27of stuff that's tedious to do by hand.
- 16:28The same animation repeated with
- 16:30different numbers—you make one,
- 16:32connect it to your data, and it spits
- 16:34out 50 versions. Think about what that
- 16:36means for your workflow. The same
- 16:37animated card with every client's
- 16:39results, the same intro with the name
- 16:41of each episode, and the same ad
- 16:43creative in 20 variations to test them.
- 16:45Things you wouldn't do by hand because
- 16:47you don't have the time: you use it
- 16:49inside your agent, let it download the
- 16:51project, and give it the brief. For
- 16:52example, make me an 8-second animation
- 16:54at 1080p, dark background, a number
- 16:56rising from 0 to 100,000 with large
- 16:58centered typography and a line of text
- 17:00appearing halfway through. Look at the
- 17:03result I showed you earlier; it’s
- 17:04truly incredible. Pay attention to the
- 17:06license here, because it's actually a
- 17:08bit unique. Remotion is free for
- 17:10individuals and companies with up to
- 17:12three people. If you are a larger team,
- 17:14you need a commercial license. Don't
- 17:16ask me why. So for you, if you're just
- 17:18starting out and are on your own, it's
- 17:19absolutely super useful. That said,
- 17:21we've reached the end. I hope these
- 17:23tools have been useful to you. I'll
- 17:24leave you with this video where I
- 17:26present other completely free tools,
- 17:27and as always, happy creating.
About this transcript
This page contains the full transcript of 10 Tool AI GRATIS Così Potenti da Rendere INUTILI Molti Abbonamenti by GabrySolution, generated from the public captions YouTube serves with the video. The transcript has 3,493 words across 512 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.