00:00There are thousands of pre-made NADN workflows out there from the official library to GitHub and other sites

00:04And on the surface, that sounds like a dream

00:08But if you've tried using them, especially if you're new to NAND, you've probably hit a wall

00:12A lot of these templates are confusing, half-broken, or missing key pieces, and it's easy to get stuck fast

00:18The good news is once you know how to spot and fix those issues, these templates become incredibly useful

00:23You can run them out of the box or use them as powerful starting points to build your own automations

00:27In this video, I'll show you exactly how to do that, plus where to find even more workflows, and how to use tools like chat GPT to help you debug or even create new nodes and workflows from scratch

00:39But to start, I'll walk through one relatively complex template from start to finish so you can see the full process and what to watch for along the way

00:46First, to get to these workflows, just go to the main NAN site, hover over product, then click on templates

00:52There are 2,990 workflow automation templates and they continually get added in here

00:58These are all community created

00:59Most are free, but if one is paid, it'll be labeled clearly

01:03You can search by keyword or category, and there are workflows for just about any use case or industry

01:07For example, under sales, you've got options like lead workflow for Yelp and Trust Pilot scraping, find and email anyone on LinkedIn, sales lead magnet agent, build a voice AI chatbot, automate business lead scraping, and that keeps on going

01:22And it's the same for other categories

01:24Marketing has templates for automated videos and content creation

01:26Engineering includes chat bots and retrieval augmented generation setups

01:30Finance

01:32You get the idea

01:32You can also search for the app you want to use

01:34So I'll try YouTube

01:36And that gives 49 different workflows just related to YouTube automation

01:41Now I do want to go with something relatively complex

01:42You know, maybe intermediate level

01:44I wanted to have multiple integrations we can work through

01:48So I will use this one

01:48Create and upload AI generated ASMR YouTube shorts with Cedance FA AI and GPT4

01:52It's probably a few weeks late to jump on that trend now, but I'm sure we can modify the prompts to fit any style

02:02When I open this up, we can see the full workflow layout right away

02:03If we scroll down, there's a breakdown that covers who this is for, what problem it solves, and a step-by-step of what the workflow does

02:12This one includes two-stage AI ideation and planning, multimodal asset generation, including adding sound effects, and then there's closed loop publishing and logging at the end

02:20Under the setup section, it shows all the required credentials

02:23So, which accounts and APIs you'll need

02:25I'll walk through APIs more later, but this one includes OpenAI, Wave Speed, and FO, which are all pay

02:32Open AI will cost a couple cents per call

02:34Not sure on Wave Speed off the top of my head, but I think that's around like 20 to 50 cents per video depending on the model

02:40and the file cost varies depending on what you're using it for, but generally that'll be cheap

02:44The rest of the services used in the workflow are free

02:47After that, it explains which nodes you'll need to manually configure and how to customize the workflow, so you're not locked into ASMR videos

02:53You can use this structure as a base and adapt it to any niche or style you want

02:57Back up at the top, I can click into the workflow and zoom around to see everything

03:00Looks like they have similar instructions laid out here, then some further instructions across each step

03:06And typically I'll look closely at what's happening to make sure it's a solid workflow I want to use

03:10This one starts with a couple open AI steps

03:12First to come up with an idea, then expand that into a structured plan

03:16The plan gets saved to a Google sheet and then turn into prompts for video generation

03:20And that looks like it will be multiple videos and sound effects

03:23Then stitching it all together

03:23It will send that off to generate the videos

03:26This wait step is because it takes a while for them to generate

03:28So it has to pause long enough to make sure they're finished before moving on

03:31Then it will get sound effects

03:33Sequence everything together in one edited video

03:35Probably just putting all the clips next to each other

03:38Usually people use a platform called Creatmate if you want automated editing that's more complex than just putting it into a sequence

03:44After that, it will update the sheets, download the final file, then it'll upload that to YouTube

03:49Then it will update the sheet that it's completed and send a notification by Telegram and email

03:52And the entire thing is set up to run automatically on a schedule

03:56So, at whatever interval I choose, it will come up with a new idea, generate all the content, assemble the video, and upload it without you touching anything

04:04It looks like this is 100% automated, which is pretty crazy that that's possible

04:10This won't be a super fancy video or anything, but a lot of the stuff that's been going viral from the most recent models doesn't have any fancy editing

04:17It's just a single clip or a few clips sequenced together

04:18This looks like everything should work, so I'll just click use for free

04:22And I'm not going to show the self-hosting setup in this video

04:25I'm just using the cloud version that is $25 a month, but you can absolutely self-host it for cheaper or even run it for free on your own machine if you've got a decent GPU and don't mind setting it up locally

04:35After clicking use, it takes me right into the credential setup

04:38This is where you connect all your accounts and input the API keys needed to use each platform

04:41If you haven't worked with APIs before, here's the basic idea

04:45They're how apps talk to each other

04:47Most of the time, they require something called an API key, which is basically a password that tells the app, "This request is coming from you." And in NADN, using APIs usually just means plug in a key, pick a few settings, and the workflow can start automating stuff for you

05:01I already have most of these platforms set up because I've used them before, but I'll walk through a couple

05:05And just a heads up, every service handles authentication a little differently, but almost all of them will have instructions either in their API documentation or directly inside NAD

05:14Usually, there's a little button that says, "Ask the assistant for setup instructions." Then it'll pull it all up and give you a step-bystep

05:19For example, OpenAI is one of the easiest

05:22You go to platform.opai.com

05:24openai.com/s settings

05:24Click API keys, then create new secret key

05:28I will give it a name

05:30Choose your default project or create a new one if needed

05:33Then click create secret key and copy it right away since you won't be able to view it again later

05:38Then back in NAN, paste that into the credential field

05:43Hit save and you're good to go

05:43These other ones in here are similar and easy to set up, so I won't walk through those

05:48And I'm actually going to skip Telegram because I don't need an email and Telegram message letting me know that it finishes

05:54So I'll end up just deleting that node

05:56Now wave and foul aren't on this list because they work a little differently

06:00I'll show those in a minute once we actually open up the workflow

06:03The credentials listed here are all filled out

06:05So I'll click continue

06:06Then the full workflow opens up

06:08It's not ready to run just yet

06:10There are a few parts we will need to fill out manually

06:11And there are a few risks you should be aware of when using community nad templates

06:15Honestly, you'll probably never run into these, but they can happen, so it's worth a quick heads up

06:20The three main things to watch out for are trigger loops that can run endlessly, exposed API keys, especially inside HTTP requests, or just suspicious HTTP nodes that send data to unknown URLs

06:31Avoiding these is pretty straightforward

06:34Just make sure triggers are set to run at reasonable intervals and don't get re-triggered

06:37Double check any URLs data is being sent to and keep your API keys from getting exposed in requests

06:43As I walk through this, I'll go into each of those a little more in the parts where they could actually come up

06:48Now, over on the left, you'll see instructions built into the workflow to guide setup

06:51Not every template has it set up this nice, but a lot do, and it helps a ton

06:55The first node is the schedule trigger, and that actually brings up the first risk I mentioned a second ago

06:59Accidental trigger loops or overly aggressive schedules

07:02Right now, this is set to run every 30 minutes, which is way too often for this kind of workflow

07:07I wouldn't want it creating and posting videos every half hour

07:08or something like this, you'd probably want to start with once per day, especially when you're triggering APIs that cost money like OpenAI or FAL

07:16You want to be intentional about how often it runs or you'll burn through credits for no reason

07:22In this case, we're using a schedule trigger, which is easy to control, but about that looping risk I mentioned

07:26Here's how that can happen

07:28So, a loop usually occurs when a trigger is accidentally re-triggered by the workflow itself

07:32For example, say you're using a Google Sheets trigger that runs every time the sheet is updated

07:36Then later in your workflow, you have a step that updates that same sheet

07:39So maybe to mark a status from in progress to finished

07:43If that update is misconfigured, like writing to a new row instead of updating the existing one, it could trigger the workflow again and again and again

07:51That's how you could end up in an infinite loop

07:53Usually, this kind of thing shows up pretty obviously in testing, and it's easy to fix

07:58But it's definitely something to watch out for whenever you're building or modifying templates

08:01These automations and AI agents we're building in NADN aren't just for personal productivity

08:07They're the same systems startups are using right now to scale fast and run lean

08:11HubSpot just released a free report called Hyperrowth Startup Index and it breaks down how the fastest growing startups in 2025 are operating differently using AI first operations, lean tech stacks, and automation built into every layer

08:24One part that stood out to me is they show how teams are turning every department, marketing, sales, support, finance, engineering

08:32They're turning those into mini tasks accomplished by many AI agents

08:34One agent finds leads, another writes outreach, another follows up

08:39This is the same logic we're building into these NAD flows just at startup scale

08:42And the report goes deep beyond just the tech stacks

08:48They use data, founder interviews, and funding and strategy breakdowns from companies like G2 and breakout startups like Clay and Goldcast

08:55There's even the bold claim that we'll soon see a billion-dollar company run by one person thanks to agentpowered systems

09:01The report is free

09:03It's linked in the description

09:05Thank you to HubSpot for sponsoring this video and providing valuable resources like this one

09:09Moving on, we have the first OpenAI node

09:11You can see the prompt it's going to send plus the system prompt which sets chat GPT's role and behavior

09:17This node is where we could easily rewrite the prompt to adapt this for different styles of videos

09:21And kind of the same thing in this next OpenAI node

09:24You can see exactly how it's building on the earlier result

09:28Then we hit the Google Sheets node where it stores the generated ideas

09:29You'll see Google Sheets pop up in a lot of animated workflows

09:33They're usually used for one of three things

09:35storing generated data, feeding prompts into LLMs, or logging what happens after a task

09:41Whenever you connect a Google Sheets node, you'll almost always need to create the sheet manually and map the incoming data to columns

09:47For this one in particular to work, I need to create a sheet in my account that matches what the workflow expects

09:53Back in the instructions, it listed the required fields

09:56So, I'll make that sheet real quick

09:58Okay, that's all done

09:58Now I go back into the sheets node and under document I'll switch to from list which will show all of the sheets tied to my account

10:05I'll select the new one, choose sheet one, then switch to map each column manually

10:10Each column I created will now appear for mapping and moving on

10:15Unbundle prompts looks fine

10:15Next up is the create clips node

10:17This is an important one to focus on for a second

10:21This step is sending a request to the wave API to generate the video clips

10:25This node and the ones using foul all work in basically the same way

10:28By using platforms like these, you can connect to all kinds of AI models directly inside your workflows

10:33But both of these platforms do use a slightly different API setup than we covered before with OpenAI

10:38They use header authorization

10:41Some APIs need you to prove who you are by sending a special token in the request headers and NAD's header off lets you do that automatically

10:47I don't want to get too sidetracked on what that is

10:50And really, you can just click ask the assistant and it will come back with the exact steps to follow, but it's easy

10:55As it says right over here, put authorization as the name

10:57Then for the value, you'll type bearer, then a space, then paste in the API key

11:02This will format it in the way wave speed is expecting it

11:05And that's all there is to it

11:07Then for file, it's the same thing except the value is key space API key

11:11So sometimes there's APIs with a little different setup, but use the assistant or chat GPT if you get stuck

11:14Or there's a custom GPT trained on NADN that I'll show in a little bit that can help too

11:21Looking at the URL in this wave speed node, it's set to use seed dance V1 light for text to video at 480p

11:26That model isn't as strong as the pro version, but it's cheaper

11:30So I'll leave it like this while testing to avoid burning credits

11:34Later, once everything's running smoothly, I'll probably upgrade the model

11:37If you go to the wavepeed site, you'll see you can run all sorts of different models through their API

11:41Juan, cling, miniax, v3, and more

11:44But for now, we'll stick with seed dance since that's what this template was built around

11:48Now, this node is also where the other two risks I mentioned earlier could show up

11:51API key exposure and suspicious HTTP requests

11:56The first is API exposure

11:56This node is using header O credentials, which are handled securely by NAN

12:00So, your API key isn't being hard-coded or sent visibly

12:04Anytime you're using stored credentials like this, you're in good shape

12:07But here's what it would look like if it were wrong

12:08If the templates ask you to manually paste your API key directly into the node, usually in like the headers or the body, like the node will work that way, but it's not secure

12:18If you were to save the workflow or share it later, that key is now embedded in the JSON

12:22The better approach, like this one, is to use credentials so the key is kept securely in NAD and it never appears in the actual workflow file

12:31Now, to be clear, I think it's pretty unlikely you'll run into a template where someone is trying to trick you with this, but it is something that's possible, and it's just good to understand how API keys get passed around

12:40The second thing is the suspicious HTTP request risk

12:42This node is sending a post request to a wave speed endpoint

12:47You can't just paste that URL into a browser to check where it goes

12:51That's not how APIs work, but you can go to the wave speed docs and confirm that endpoint matches a real model

12:57When you're reviewing HTTP nodes, generally this is more likely to be a concern in templates from like GitHub or unknown sources, which I will cover later

13:04Basically, just check what is this sending, where is it going, does it include any sensitive data

13:08Again, this is probably not something you'll have an issue with, but it's just good practice to understand what each node is doing, and where your data is being sent

13:15You don't need to know every line of JSON, but at a high level, I should at least know that this node is sending the video prompts to Wavespeed, asking it to generate the videos

13:24In this case, everything looks clean

13:26It's a wave call

13:28It's using stored credentials and we know what model it's running

13:30After that, we hit the pause node while the clips generate

13:33Then it gets the generated clips

13:35Then we move into sound effect generation

13:37This step is set to use foul, but it defaulted to my wave speech credentials

13:42So I'll switch that over to foul

13:45Same thing with the get sounds step

13:47That also needs to be switched to foul

13:52Then in sequence video and get final video, same deal

13:54I'll switch both of those to foul as well

14:01After that, we hit another Google Sheet step

14:03So, just like earlier, I'll pick the correct sheet and map each column manually so everything lines up

14:06Then, it downloads the final video using the URL stored in the sheet

14:10After that, it uploads directly to YouTube

14:13In here, I just need to set up my region code

14:17Choose a category name

14:17Doesn't really matter what, and it shows a description that'll be added

14:21I'll leave that as is for now

14:23Then final step, another Google Sheets update

14:26Same process as before

14:29At this point, the whole workflow is set up, but there will still be some errors once I test it

14:33And whether I'm building a workflow from scratch or starting with a template, I usually have ChatGpt open to help troubleshoot anything that comes up

14:40Most of the time, you can just drop in a screenshot of the error, and it'll give you a step-by-step fix

14:44Now, personally, I use a custom GPT for this

14:48Now, I didn't create it, but if you search NADN in the GPT store, it's the most downloaded one

14:52I haven't tried the others, so maybe those are better

14:54I don't know

14:56But this one includes all the current NADN documentation, so it's usually more accurate than just the default chatbt

15:01But when I'm using it, I usually start with a prompt like I'm using an NAN template

15:05I need help setting it up, checking for errors, and identifying any security risks

15:08Here is the full template

15:10Then back in NAD, I will highlight the entire workflow, hit command C to copy, then come back to chatbt and paste

15:17that will paste in the JSON of the entire workflow and chat GPT will understand exactly what it does

15:23And this GPT asks your experience level and how you want the explanations

15:26And now it gives a full breakdown, a highle overview, what's working well and any issues it finds

15:32In this case, it pointed out a few simple problems

15:34The Telegram error is just because I never added my Telegram credentials

15:38Then there's an extra space in a URL and I forgot to add the email address for the final notification

15:43Easy fixes

15:43I will remove the space

15:46Since I don't need both email and Telegram notifications, I will delete the Telegram node

15:50Then I'll go into the final email node, add my address, and that's good

15:53This is close enough now to just run a test and debug as we go

15:57So I'll click test workflow, and the open AI step runs fine

16:01But now I've hit an error in the Google Sheets node

16:05I actually left this part unfinished on purpose

16:06It's easier to handle once you see the incoming data

16:10The error is because the values to send section needs to be filled out

16:11But I'm also noticing that the incoming data has different fields than the column names I originally added to the sheet

16:17I went off the ones in the instructions, but I don't think that was accurate

16:20And this is all on the left side in this input panel

16:23That's the data coming into this node from the previous step

16:25I'll go update the Google sheet to make sure there's a column for each of those fields

16:30Then I'll come back into the node and click refresh fields

16:31And now the new column names show up properly

16:35And these values need to be in JSON format, but you don't have to write any of that yourself

16:38You just drag each field from the left panel into the matching column and then NADN will format it for you

16:44So I'll do that for each one

16:45Then I'll click test step and make sure it works

16:47Over on the right side now we'll see the output panel that shows what's being sent to the sheet and also passed along into the next node

16:55And everything looks accurate

16:55Now I'll just hit test workflow again to see how far it gets

17:05It looks like it actually made it all the way through the next OpenAI step

17:08then generated the videos, added sound effects, and even stitched everything together

17:12The only failure was when it tried to add the URLs to the Google sheet

17:15So, we're really close

17:15I'll open up that step

17:17And yep, the incoming data includes a video URL and thumbnail URL, but I didn't create columns for those yet

17:24So, I'll go ahead and add those columns to the sheet

17:25Then, I'll just match this on row one

17:27That should work for now

17:29Then, I will drag over the values from the input section and hit test step

17:37And this time it worked

17:37The spreadsheet now shows the video and thumbnail URLs

17:41Except I just realized I overwrote the header row since I used row one for the data

17:45So I'll need to fix that and set it to start writing below the headers

17:48Next, I'll look at the node that failed right after

17:50That's the YouTube upload step

17:52There's an issue though

17:52So I'll take a screenshot, drop it into chat GPT, and have it help me debug

17:56The problem is that the title field isn't showing up in the input

18:00It needs to be pulled from an earlier step

18:01Chat GPT spotted that and gave me the JSON to pull it from the sheet

18:05So I'll paste that in

18:06This is honestly the most important skill when working with NAD templates or NAD generally is debugging small issues

18:12Even great templates will break if they're just missing a credential or if one field doesn't match what your sheet expects

18:18Getting used to testing nodes one at a time, reading the input and output panels, and pasting into chatbt when needed

18:23That's kind of the whole game

18:25Now, I will test this node again

18:27And it hit another issue

18:29All right, I'm skipping ahead here because this one took a little bit to figure out

18:32It was a weird error

18:32So, what it seems like is this category doesn't exist on YouTube anymore

18:35So, I just had to switch it to a different one

18:38But the process of figuring that out took more steps than it should have

18:42So, I'll spare you watching those boring back and forths, but that will happen sometimes

18:45A weird random issue that's hard to figure out

18:47Honestly, that is just part of the process

18:49Expect that

18:51Anyways, after switching categories, I ran it again and this time it uploaded successfully to YouTube

18:55So, I can go check there

18:57Looks like it's live

18:59That's perfect

18:59So, we can watch it really quick

19:24Everything came together, although I'll definitely want to upgrade the video model

19:28Then I just need to fix the final sheet step to update the status to posted

19:31Then this entire workflow is complete

19:34It should run end to end all in one go

19:36Before I run the full workflow again, you probably noticed right there that the video quality isn't great

19:41That's because it was using the light model

19:42So now I'll switch it to the pro model for this final run

19:44I'll head back to that node, then over to the wave speed site to double check the exact way it's named

19:50And I'll change that in the node and everything should be good to go

19:53Then I will delete the previous execution data and test workflow

20:03So after a few minutes, it says it is all complete

20:05So I'll go check my email

20:07There it is

20:07Now let's watch the final video

20:41Okay, that yeah, looked good

20:41A lot like some of those AI ASMR videos that went viral a couple weeks ago

20:45It's not 100% there yet

20:47I'd probably test out some other video models, refine the chatbt prompts, but really I'd try a completely different niche altogether

20:53If this is something I was into longterm, I'd probably build something that scrapes high performing videos, analyzes them, and uses that data to generate smarter prompts automatically

21:01That's not the point of this video

21:03For this demo, this workflow is solid and fully automated end to end

21:07And once you get one of these working, you start to realize how powerful NAD templates are

21:10They're not perfect, but they can cut your build time by 80%

21:14And once you know how to test, fix, and remix them, you can build some pretty advanced automations without ever starting from scratch

21:20You should definitely also learn how to do everything from scratch

21:24I do have an entire video teaching that, but this can save a lot of time

21:28I want to show just a quick example where I have chat GBT generate a full node from scratch

21:32And here's one where I wanted to replace a couple of nodes

21:36And then asking chatbt to do that can just help save some time

21:39And this is a YouTube comment sentiment analyzer, which is actually a pretty useful idea

21:43So like scrapes all the comments on one of my videos and tells me what the general sentiment was, what types of questions people were asking, things I could do to improve, that type of thing

21:51So, I was going through this one and I was having issues with the OOTH YouTube API

21:54I just decided to switch it to a simpler API to use

21:56So, I just asked Chatbt to replace the nodes with HTTP request nodes using my other YouTube API that I already had set up

22:04It needed to replace all three of these nodes, which would take a few minutes for me to set up on my own

22:08But with Chad GBT, I just asked it to give me the full JSON for them

22:11So, here it is

22:11I was already in a chat, so it knew what I needed

22:15Then, I just gave it a couple more details, and that's it

22:17It just asked it for the nodes and gave me the full JSON for them

22:20So I take that and paste it in here and I have all three nodes

22:24So I just remove the others and connect it to the new ones and fill out the fields

22:28Then the whole workflow should work now

22:30So that can definitely be a timesaver

22:31And if you get really detailed on what you want a workflow to do, you can even have chatgpt generate the entire workflow, every single node, all from scratch

22:39Just from one prompt that you give it, it will generate the entire JSON for a full multi-step workflow

22:45So I wanted to cover that

22:45It does feel kind of magical when it works, although it doesn't work every time

22:48But now that you're familiar with how all of these templates work, let's look at a few other places that you can go to find even more workflows

22:55One of the best sources is GitHub

22:57There's a few large repositories that collect free workflows from all over the internet

23:01One of them is this one called Awesome Nad Templates

23:05It's organized into categories, and there's tons of workflows in here

23:08I'll try Gmail and email automation

23:09Then how about a email summary agent? And when I click into it, you'll see it's just a raw JSON file

23:16And you can use that in a couple different ways

23:18So, first you can just click copy raw file up here

23:19Then in NADN, create a new workflow and paste that in

23:24Just like that, the whole template is imported

23:26This one's fairly simple, but it includes notes on every step, which is great

23:29Not all of them will be this well documented

23:31Then the second option is to download the file from GitHub

23:34So, I'll download that

23:34And now I have a text file

23:37I can either open that and copy the JSON just like before, or in NAN, I can go up to the three dots, then click import from file

23:46I select the file from my computer and that will load it up as well

23:48This workflow looks solid, but with these that we have in GitHub, there's a ton that are just scraped from all over

23:53So, there will probably be more workflows that are outdated or have issues

23:56So, that first step of having chatbt look through the whole workflow to spot issues is extra helpful here

24:01And there are thousands more workflows available through these GitHub repos

24:05So, there was this awesome NAN templates one

24:07And there's another similar one

24:09This is called NAND workflows

24:10It's from zie 619

24:13There's a ton in here

24:13Another is from what's up J called Naden free templates

24:18This one is a curated set of 200 workflows

24:21Says they're all plug-and-play

24:22You import the workflow, add your credentials, hit activate, and you're live

24:26I just looked at a few

24:26I didn't verify all of these, but they looked solid

24:29Just keep in mind none of these are guaranteed to work out of the box

24:33They're starting points

24:33But once you know how to import, test, and tweak them, these repos become a gold mine

24:39Whether you're finding workflows from the main NAND template library from GitHub, or you may come across them on YouTube or X, everything I've covered here should give you what you need to jump right in

24:48Once you know how to test, tweak, and fix the common issues, you can use just about any of these templates to start automating whatever you're working on

24:56So hopefully that's helpful

24:58And if you want to go way more in-depth learning AI on Futuredia, we have over 20 comprehensive courses on how to incorporate AI into your life and career to get ahead and save time

25:06These cover everything from content creation to design to coding

25:10You can get a 7-day free trial using the link in the description or watch this video for a deeper dive into the specifics of NAND and building workflows from scratch.