← Video Library · ChatGPT & GPTs

Setting Up a Custom GPT from an Existing Prompt in ChatGPT

Published December 2, 2025 · 323 views on YouTube

Key takeaways

How Do You Turn an Existing Prompt Into a Custom GPT in ChatGPT?

You create the GPT from your profile menu’s “My GPTs” screen, then use the Create chat to upload your existing prompt or workflow as a knowledge file. ChatGPT’s builder analyzes that content and drafts the name, description, instructions, and conversation starters for you, which you then refine conversationally.

What Are the Steps to Build a Custom GPT from a Prompt You Already Have?

  1. Click your profile icon and select “My GPTs” to open your GPT list.
  2. Click “Create” to open the GPT builder’s conversational interface.
  3. Upload your existing prompt or workflow (for example, a TXT file) to the knowledge base so the builder trains on it.
  4. Paste the same content directly into the Create chat as additional context.
  5. Decide whether to enable extras like web browsing, DALL-E image generation, or code interpreter (skip “Actions” if you don’t need external URL/API connections).
  6. Click “Create the GPT” and let the builder draft a name, description, instructions, and conversation starters from the uploaded content.
  7. Answer the builder’s follow-up questions (tone, style, audience, level of creativity) to refine the configuration.
  8. Test the GPT in the preview panel on the right side of the screen.
  9. Save the GPT, give it a name, and choose a sharing option: only me, people with the link, or public.

What Knowledge Base Content Works Well for a Custom GPT?

Any long-form reference material you want the GPT trained on works: a full book manuscript, a script, a customer-support knowledge base, or — as demonstrated in the video — a complete “hero’s journey” recipe covering all 12 stages of the framework. The builder reads the uploaded document, outlines its structure, and uses it as the basis for the GPT’s instructions.

Can You Adjust a Custom GPT’s Tone and Personality After Building It?

Yes. The builder keeps asking clarifying questions during setup, such as whether the GPT’s approach should be more traditional storytelling or more innovative, and what tone (formal, casual, or a mix) it should use. Each answer updates the “Configure” tab in real time, and you can continue tweaking instructions, the knowledge base, and the profile icon (including generating a new one with DALL-E) at any point.

What Happens After You Save a Custom GPT?

Once saved, the GPT appears in your GPTs list under the name you gave it. You can open a direct chat instance to test it, and return to “Edit” at any time to modify the instructions, knowledge base, or description. In the video’s example, a “Hero’s Journey Builder” GPT is tested by describing a fictional character (a hobbit named Sam who needs shoes), and the GPT walks through the hero’s journey stages — ordinary world, call to adventure, refusal of the call, meeting a mentor, crossing the threshold — applying the uploaded framework to the new story.

Full video transcript

Hey y’all, Darby here with a video on how to set up a very simple GPT from an existing workflow or recipe. Our hero’s journey recipe is going to be the example demonstration here today.

I’m going to move over to this side, and inside of my ChatGPT interface, you can see up here in the upper left-hand corner these are the current GPTs that I’ve made and been messing around with. You go down to your icon, you can select yourself, which will open up this folder. You go to “My GPTs.” That will open up a new screen and we’ll move into the next step of creating your GPT.

So you can see here, here’s my GPTs. You can see some recently used, how often it’s been used in conversation, some by OpenAI, and I’m sure this will change as time moves on. But now when I click “Create GPT” up here, it’s going to open up this interface so you can talk and just chat with the GPT builder, and you can configure name, description, instructions, conversation starters, and so on.

What I’ve done is I’ve actually downloaded a TXT file of the hero’s journey recipe, which I am going to upload right here to the knowledge base, and this is me training it on a very particular workflow and outcome. So this is going to be the hero’s journey. You may have a book if you’ve written a book yourself, or you’re looking to reference materials like scripts or knowledge bases for customer support or a number of different things that have a lot of content related that you want to train up on. That’s where this knowledge is going to go.

You can see web browsing, DALL-E image generation, and code interpreter are options. I’m not going to select code interpreter yet. Then you have a few ways to add additional settings, like “use conversation to improve our models,” and I’m actually going to say no on that one. And then “Actions” is going to open up an ability to tap into other URLs, and I’m not going to get into that during this video.

What I am going to do, now that I’ve uploaded my knowledge, is go to “Create” and I’m going to paste the same contents I’m training it on, which is the hero’s journey. And this is how the process will get started. As you can see, this is the entire workflow — one version of the hero’s journey, all 12 stages of it, that we broke down into a giant recipe. And now ChatGPT’s builder is outlining this document. It’s understanding, okay, this is what we’ve uploaded, and it’s analyzing the stages and doing all of this itself. So whatever you’re training it on, whether it’s one specific prompt or how to answer questions, understanding your customer service — okay, great, let’s create the GPT.

Now we’re just feeding it the information and using it as context to craft what’s ultimately going to be the configuration of this GPT. You can see here the GPT builder is going and configuring — there’s nothing here right now, but that will start to get ironed out as we build the GPT. It’s going to come up with some outputs related to the context we’ve fed it and the instructions we want it to follow, and create these types of outcomes.

Again, this being the hero’s journey, naturally, whether you’re creating a fictional character or a customer journey for somebody within your business, there might be a few different ways you can utilize a framework like this — from storytelling to writing scripts to just brainstorming different relevant ideas. This GPT will help focus in on the specifics of the customer journey.

This is an example of it being built into action. You can generate logos and icons with it — this is a neat little open-book-looking icon with a bunch of different cool art inside. Next, “ready to guide through the 12 stages of the monomyth, offering prompts, questions, guides. The profile picture represents a GPT of a mystical book opened.” Cool, we’ll keep it like this for now — we can easily change it, use DALL-E 3 to change it, or upload your own icon. So you can do whatever you want there.

Then we’re refining the context: “How would you like it to approach — traditional storytelling or innovative?” I’m going to say a mix of both, keeping in mind people use this for fiction, storytelling, customer journey, and more. We give it that information, and now it’s building and refining. I go to “Configure” again, and now you can see a guide for character and story development using the hero’s journey — this GPT is designed to assist with tools to enhance the storytelling process. That’s the instructions, and we also had conversation starters show up automatically. You can add more if you want; I’m not going to, so I’m just going to delete one.

So we’ve got these here and we’ve got the knowledge base, and it’s pretty straightforward from there. It’s still updating — at this point we can start to test and see how it works, and I’ve been noticing it’s timing out here and there depending on the generation. It’s kind of interesting to see how fast this is going to develop and maintain uptime — if you’re building and using these tools internally for your business, you’re going to want this to be working as often as possible. It’s trained itself with GPT-4 and a blend, and I’m sure OpenAI is investing a lot on their end in this technology, so it’s only going to improve.

More questions come up: tone — formal, casual, or a mix of both; include the user’s own perspective; scale of imagination for the AI to build off; be conversational and vivid using emotional language. I’m just responding back and carrying on the conversation, which is informing the GPT how to be configured. Then we can ask certain things — I’m going to say “hi,” and it walks you through the process, saying it can assist you with your character today, going directly toward the storytelling aspect. It also asks about personalized interactions based on the user’s level of experience — I say yes, that’s good.

As it’s updating, we can continue to play around with it, and now it’s basically fully personalized. So we can test it out in the playground. I’m going to save it. Once you save it, you have a few options: publish only to me, people with the link, or public. Mine is going to just be only me and people with the link.

So I’m going to go back to my chats, and there we see it’s still untitled, so I’m just going to give it a name — “Hero’s Journey Builder” — and save it. Now anytime I want to build, I would just go to edit and modify and tweak and update the descriptions. I just deleted that and it went away from over here. So these are all prompts you can use within your own. Then, of course, you can open up a direct instance of it in your GPTs.

Let’s confirm — and so here we are in the Hero’s Journey Builder, and it has all that stuff. So: “Hi, my character’s name is” — let’s just say Sam — “and you’ll start the beginning of the story.” It asks, “Could you tell me about Sam’s background, the world he inhabits?” I say Sam is a hobbit in the modern world who needs a new pair of shoes — we provide shoes for hobbits as a brand, this is their customer journey. Totally made-up example, obviously based on Lord of the Rings, with Sam in the modern world — my company makes shoes for a brand that makes shoes for hobbits, because hobbits have big feet and don’t wear shoes. Probably a terrible business idea, but you can see here now we’re going to the ordinary world, the call to adventure, the refusal of the call — he might be hesitant, unsure about modern brands, the cost involved — meeting a mentor, where the brand’s messaging resonates with him, and that brand becomes the mentor — crossing the threshold.

So now we’ve basically built in this hero’s journey framework for it to reference, and it’s creating that story arc based on the context we’ve provided. Whether you’re making your own fictional character or a character for your own brand, you can utilize this to help build out your hero’s journey and use it as a map to help your customers or characters hit all the different beats within that story using this framework.

So that’s the basic example of how things get started, and then you could go back and forth with it to keep things going. Other than that, very simple to get a basic GPT up and a lot of fun to play with. So dive in, let me know how it goes for you, and see you in the next video.

Watch on YouTube

Related tutorials