AI versus Marshy #52 - fashion, picture editing, and building my biz


Hello and welcome to another edition of AI versus Marshy đź‘‹

This is the newsletter that is excited about AI, but keeps it grounded in things you need to know and can do today.

This week we look at:

• Using ChatGPT for fashion advice

• Editing pictures with words

• Another build in public update #2 - TheLeadMagnet.biz

Lots to share with you today, so open like a door and make room.

-Marshy

​

ChatGPT for Your Own Wardrobe

Via Ruud Hendriks on LinkedIn.

My friend Anna shared this practical example for people who are less than fashion-minded (people like myself!)

Ruud wrote up on LinkedIn an example of using GPT Vision to recommend items of clothing.

I got excited by this one and bit off more than I could chew.

<< warning - NERDING OUT section - skip the next few paragraphs to avoid nerd >>

One of the things I learned recently is that you can use ChatGPT to create "synthetic data".

This is a fancy way of saying dummy information.

This is useful when you're building things as you need data to feed it.

For example - you might be prototyping a client analytics dashboard, but you don't have live data to feed it.

Asking for dummy data can help move it forward until you get access to live data.

It turns out you can also do this with image generation.

Well... to a point.

I wanted to build my own GPT for this but wasn't near my wardrobe - so I asked it to create a version of my wardrobe for me.

​

It was being lazy and trying to put them into one image, and this didn't suit me as I wanted separate images.

​

​

Then I hit another unexpected snag - an image generation limit!

You're generating images too quickly. To ensure the best experience for everyone, we have rate limits in place. Please wait for 13 minutes before generating more images.
If you have any other requests or need help with anything else in the meantime, feel free to let me know!

I ended up editing some of the wardrobe with an image editor and then decided to put the wardrobe into a database.

I created a form for uploading the "clothes" and had a database I could "call to" from a GPT.

This is where things got tricky.

Last edition I mentioned GPTs let you create "Action" buttons.

These are ways to pull data into and out of a GPT.

So I created my own GPT.

And started trying to work out how the Action button works.

I wanted to to look up items in my Airtable database, and decide on what I should wear based on my prompt.

This is where I hit the wall of my own understanding.

It required:

• Authentication - I can do this part

• A schema - okay I'm moving into unknown territory

• Running scripts in python based off this schema and connecting it to Airtable, OpenAI API Key, and working - alright we're just completely freeballing here

An amusing (but ultimately unsuccessful) back-and-forth process ensued where I would hit an issue, troubleshoot with ChatGPT, get past the issue, then circle back.

​

I got the action button set up, but couldn't do anything with it.

​

​

<< end nerding out section >>

So that's why the newsletter is late.

I spent my writing time trying to build something beyond my capabilities!

Funnily enough, I can just upload the images into ChatGPT normal style (it takes 10 at a time) and then suggest what I should wear to a startup mixer.

Spiffing! 🤣

Telling the picture to change with words

Back in edition #3 I reviewed Photoshop's Generative Fill feature.

That was a year ago - and in a surprise to absolutely no-one the design industry is still doing fine.

Photoshop is an expensive and high-end tool that requires skill to get the most out of it, and that still applied to the generative fill feature.

But since then ChatGPT 4o has released a feature that I actually haven't seen a hell of a lot shared about - which is the image editing feature.

You can find it when you click on an image you've generated you can make amends.

So let's play!

Here's my latest professional headshot:

Let's turn it into a cartoon.

​

I have a button now that let's me highlight sections and make changes.

I don't think it's a generative fill replacement, but being able to tweak things to showcase what you need a bit better could be helpful.

I often get asked about what to do, what apps to use, or what's the best thing to do right now.

I believe it's simply playing and seeing what you can do with it.

It's the best way to learn - and while the first example managed to stump me and delay my newsletter - you can bet it's not going to prevent me from figuring out something similar in the near future.

​

Sprint 2 - TheLeadMagnet.biz

For new readers - my goal is to build a self-sustaining business that I can run solo and use AI to reduce the heavy lifting.

Last week I soft-launched the website, had one solid enquiry, and it hasn't gone much further.

I got held up waiting for my old FB Page called "Marshy Idea" to rename to TheLeadMagnet.biz - in the end it just wouldn't rename, probably to do with it being a 10+ year old page.

So I created a new FB page (for ads) and created 2x landing pages with Carrd.

Landing pages are a core part of running ads in 2024.

They let you test and tailor messaging based on the different kinds of ads you run, and improve things over time (known as A/B testing).

ChatGPT helped me by suggesting 2x ad variants, and 2x landing page variants.

The former were bland, but the latter was good - giving me a structure and outline for a detailed, in-depth page, and a short and direct page.

I was able to refine these and get them live:

Web versions:

​https://lp1.theleadmagnet.biz/​

​https://lp2.theleadmagnet.biz/​

I was able to grab testimonials from way back when I tested speedmarketingdemos but am after more.

I've now got my page ready to go and will set the ads live at $30 a day.

​

I'm expecting minimal results here and am banking on the power of the sprint and learns.

Each week I can cycle through Sprint 1, 2, through to 20 and know that things will improve with consistent effort.

The goal is to build a system I can sustain - the traction will follow.

I've been wracking my brain thinking about a "home" for building in public with this project, but I think it's a good fit in here and look forward to sharing about next week's sprint!

--

Last night I was on the couch and said to G I'm tired.

She said "yeah no wonder".

Saturday - facilitated a 1-day workshop for Startup Gippsland in Mirboo North

Tuesday - off to Canberra with SANE lobbying for more funding

Thursday - attended the Digital Day Out - a local conference with business owners

So I'm looking forward to a relaxing weekend.

Thanks for reading and if you've got a question about AI you want to talk through drop a time in here.

Happy to help with no-strings.

We've got this!

-Marshy

​

Early Growth School

Continuous tips, lessons, advice, and direction on growth for your early-stage startup by a growth marketing pro.

Read more from Early Growth School

how to grow a hiking app. I spoke to a founder with a hiking app today. I don’t normally work with pure B2C founders but I just couldn’t help myself with our meet and greet. “Sounds awesome - do you want some ways to get more users?” “Well yeah” The app in all its local glory Here’s what I told him: Do you know who loves nature walks? Older people with more time on their hands. Nearly every suburban area in Melbourne (the area we’re in) has a community learning centre/neighbourhood learning...

Let's flipping go! You're getting this email because you're a previous subscriber of the now obsolete AI versus Marshy. In case you missed it - here's the farewell post. You're now getting the new version of the email - early growth school - which is designed to give early-stage founders clarity around their growth efforts with continuous tips, lessons, direction, and advice on what to focus on. If that sounds terrible and not up your alley - please click this DNW button - it is absolutely...

Hello and welcome to the last edition of AI versus Marshy! It feels weird writing that sentence - but it also feels very right. Let’s rewind to a couple of months and explain why. Have you ever had a niggle that won’t go away? Well a couple of months ago this niggle was in full flight. Here’s the niggle: This current wave of generative AI isn’t good for the planet And once I started digging into it I couldn’t unsee it: Google renegging on their carbon emissions goals - their greenhouse gas...