New Post

Happy Monday - Issue 1

Wow, I was surprised at the response of the newsletter announcement. I guess I shouldn’t be every week I have been getting messages asking when I was going to start publishing more content besides what I was posting on Facebook.

One of the benefits of bad weather in the midwest is you spend more time inside working on projects…

This weekend I was able to dive deeper into one of the AutoGPT options, GodMode.

This is a free tool that gives you the ability to assign tasks and goals to ChatGPT and through a browser interface, it determines what it needs to do and automatically executes the actions needed trying to reach the goals you have set.

I think I may have set the bar too high for a first-round test for this one and need to revisit it. I asked it to write a SEO-optimized article on a topic. Besides being able to determine its own next steps, it also has access to the internet so can source information outside of the language model.

As it went through the process, each step of the process came back with a verification from me to proceed with its next step. This, and running through a browser interface made the process extremely long.

I’ll give it a few more tries today and include a video in tomorrow’s issue so you can see it in action.

Second weekend project, solidify my process of automating YouTube video transcription for a version of the Gary V model I showed on Facebook a couple weeks ago. So far, the best out-of-the-box solution has been Shipyard, a very technical Zapier or Make. Out of the box, it doesn’t include timestamps so not a complete solution, but it is able to automatically download the audio from a video to transcribe.

Eventually, I think I am going to build a custom solution for this, but I was able to improve the manual process automation with MacWhisper.

MacWhisper is an incredibly inexpensive solution with no monthly code to transcribe and translate any audio in multiple formats. Basically, you are bringing OpenAI’s Whisper technology down to your Mac including the language models.

I am running on an M1 Macbook with 8GB of memory and transcribed a 34-minute video in 2 minutes 41 seconds for free.

Depending on the length of the video, I need to add tokens to split the transcript into multiple parts, which is currently another manual part of the process.

The cool part of MacWhisper, like Vidyo and Munch, you can simply copy paste the video url from YouTube into the software to start the process.

In the near term, I am staying manual and creating a Trello task anytime a new YouTube video is added to my client’s channel.

Well, that was my weekend. What did you do this weekend?

Watch your inbox for tomorrow’s issue where I’ll be sharing some walkthrough videos along with Plask, and tool that uses a 3d model to pose your subjects how you want.

Keep discovering,
Kevin