I've had many times where I know ChatGpt can help me with something, but I can't think of the best way to ask. So, I made a solution that gives me the perfect prompts every time. Another issue I run into is trying to get the words right for generating images or videos.
Maybe I know the concept, but not every detail or style. Or sometimes I'm just stuck completely. I made a solution for that, too.
I'll cover how I did it so you can create this for yourself, customized exactly how you want. Then, you'll always have the perfect prompt for every situation. What we are building here are prompt enhancers.
I'm going to start with text to image and text to video because that's just a lot easier. I'll cover LLMs after. There's a lot more ways to customize that because everyone uses them for different purposes.
Starting with text to image, the goal is to be able to input my basic prompt with whatever details I have in my head. then it will give me back a more complete prompt with any missing details and formatted in an ideal way for image generators like midjourney flux or any of the others. Now I have this broken down into three parts.
The core purpose, prompt structure and tips, then prompt examples for training. The purpose lays out a highle overview of what I'm looking for and how I want it to respond. It assigns the role of prompt enhancer, explains its purpose and some of the basic most important tips.
Also tells it only to respond with the enhanced prompt so it's easy to copy and paste. I did put in there to use best practices for AI image generation because LLMs are already trained on that, but I also have a big list of the specific tips I'm looking for. Now, some of these I got from Midjourney's documentation on prompting tips.
A lot of it is tips I've curated. I made an hourlong midJourney master class a while back where I covered a lot of these. This is a little long, so I won't cover the whole thing, but it has examples of how to be concise.
Instead of show me a picture of lots of blooming California poppies, make them bright vibrant orange and draw them in an illustrated style with colored pencils. Instead of that, use colored pencil illustration of bright orange California poppies as focus on what you want, not what you don't want. Balance directness and detail.
Some of this was in the purpose section, but it includes examples here. And I have the prompt structure that starts with a general overview of the scene using recognizable archetypes and keywords, subject details, setting details, then style details. Then I expanded on that with various examples underneath all of those.
Then the final section is example prompts that use the proper structure and formatting. Including examples is called fshot prompting in the prompt engineering world. Providing examples of how you'd like the LLM to respond can greatly improve the output across all sorts of applications beyond what I'm covering in this video.
I provided 12 different prompts across various scenes and styles. So that's the whole thing. It's obviously easier to explain than it was to write all this out and fine-tune it.
So I will link this in the description if you'd like to use it. I encourage you to modify it to your own preferences, though. But it's a good base that should save you a lot of time.
So I'll show the different ways to use this. Starting with the easiest, just using it like a prompt. So I have a new chat here.
I will just take this entire thing. I will copy that and then paste it into the prompt box and then send that. Understood.
Please proide your initial prompt or idea for enhancement. So I can just type in a concept, a wizard casting a spell. Send that and then it should respond with the enhanced prompt.
A bearded wizard in a flowing deep blue robe stands at top a windswept cliff. One hand raised as he casts a glowing arcadan spell into the stormy sky. Magical runes swirl around his fingertips, crackling with blue and violet energy.
His eyes glow with power beneath a wide-brimmed hat, and his staff pulses with light. Dark clouds churn above, lightning flashes in the distance, dramatic fantasy lighting, highdetail digital painting, cinematic composition inspired by classical fantasy art. So that's a much better prompt.
Then of course I could refine it further from here, but let's try another one. A robot in a forest studio Giblly style. A gentle mosscovered robot stands quietly in a lush sundappled forest clearing surrounded by tall trees and glowing wild flowers.
It goes on from there. You get the idea. That looks great.
And I will almost always take these and modify them a little more from here. But this helps a lot. You can also ask to respond with multiple ideas.
Then you can mix and match the best bits from each prompt. I built that into the core instructions of the text to video prompt enhancer that I'll cover next. I'll test these out in Mid Journey really quick.
And so you can see, of course, this is midjourney, so it's going to look cool every time, but the first version has just a wide variety and doesn't really know exactly what to do with it. The second version is a lot more fine-tuned and all with a more specific pose, so it's a far more controlled scene. And same with the robot.
All four of these are quite a bit different. Then when we scroll up, they're all a lot more similar because it knows exactly what we wanted from it. This method will work in any LLM.
Could be Claude, Gemini, anything you want. Then you just come back to the same chat anytime you need. But I prefer to do this as a project or custom GPT.
I like to have both options. Actually, I will start with a project and I'll just stick with chat GPT, but projects work basically the exact same way in Claude. So I go down and create a new project.
Give it a name, text image enhancer, create project. Now I click add instructions. Then I'll go back and copy this first paragraph.
Paste that into the instructions. Save that. Now I'll click add files.
And I created this file that has the prompting tips and tricks and it also has the example prompts for training. So I'm just going to add that file in here. So I've got that and now it's ready to go.
I just type in my basic prompt. A banana wearing sunglasses on roller skates pop art. There we go.
A much better prompt. These projects always stay up at the top in this sidebar here so they're easy to access. Then it keeps each of the chats organized.
So every time I start a new one, they'll all pop up under here. way better organization. Then the other way to do this is with a custom GPT, which has pros and cons, but I'll show you how to do that really quick.
Just click on explore GPTs, then create. And I'll do this custom. So I'll click configure.
I'll call it text to image prompt enhancer. Type a short description. Now I'll paste that same paragraph into the instructions section.
Then in the knowledge section, I click upload files and then upload that same document. And I can have Dolly create a quick image for this. That's all there is to it.
I'll give it a quick test. A dancer under the moonlight. Kirao.
And there it is. Although that one was a little long. I can click create to finalize that.
Then I can go up here and make sure it's pinned to my sidebar. That way I can access it whenever I want. I like to use projects most of the time so I can find previous prompts if I need them.
But it is nice to have a GPT as well because anytime I'm in an existing chat, I can type at then I'm able to select that GPT from the list. Now, when I type a concept, it will use that GPT to create the enhanced prompt in whatever chat I was already in. I'll still primarily use it in the project.
But once the instructions are already created, it's just easy to copy and paste it into a GPT, so might as well. I used a lot of prompting techniques that I've learned over the past couple years to help with writing these. I learned those through a lot of trial and error and reading through pages and pages of guides and documentation.
If you want to learn those prompt engineering concepts in a much easier way than I did, there's a free resource down in the description provided by HubSpot, who sponsored this video. It's titled AI Prompt Engineering Quickstart Guide. It covers a three-step framework for crafting perfect prompts.
As I've been covering in this video, the quality of Chat GPT's output largely depends on the quality of your prompt. This guide will help you master prompt engineering to get consistently great results. It breaks down the three-step framework, then gives some power tips for better results.
It also includes a troubleshooting guide if you're still struggling. Learning to communicate with AI is a valuable skill that everyone is in the process of learning right now. This guide can help you speed up the process.
Again, that's a free download using the link down in the description. And thank you to HubSpot for sponsoring this video. Now, I'll move on to text to video.
Now, this process is the same, but the instructions are different. I'll highlight just a few of the differences. A big one is that the prompt structure is different.
That's because for video, it's important to focus on the camera movement. And in addition to my own experimentation over the past couple years, I primarily drew from Runway's documentation on prompting, there were a few other places, but Runways was the most helpful. Now, for this one, I added into the instructions to respond with three enhanced prompt options.
I prefer that with text to video. Then I can pick the best prompt and also use any specific parts I liked from the others. Then there was some prompt writing guidelines and I have all the example prompts below just like before.
So I can copy that and paste it all into a new chat. Now I'll give it a test. I'll use the same concept as before to see how it changes.
Right now, there's three separate options that are each much better than the one before. And I'll make a quick project again, too. Name it, paste the first part into the instructions, and upload the knowledge base into the files.
Then, it's good to go. I did also create a GPT for this just using the same steps as before. Now, I'll try running some of these prompts through cling.
[Music] Now I'll move on to the LLM prompt enhancer which is like training Chad GPT to prompt itself. I'm going to show a general purpose prompt enhancer that detects what type of thing you're looking for and adjusts its responses accordingly, but many people will have a specific use case they come back to chat GPT for regularly. You can also create a more specialized version of this.
I'll show how to do that as well. Everyone is going to have a different purpose for theirs, but I'll show a detailed example that I personally use. But starting with this general purpose one, I'll go through this section by section explaining the logic here.
So, I gave it a role and explained its purpose. The important parts of this are expanded on below. This goes through multiple steps which encourages it to use chain of thought reasoning.
Step one is determine the best refinement approach. Analyze the user's request and determine the most effective way to improve clarity, specificity, and structure. Choose the appropriate approach based on the prompt type.
So, it needs to figure out the best approach itself and adapt. Asking to summarize an article is going to warrant a very different response than creative brainstorming or multi-step instructions. So, I gave multiple examples of how that would look.
An important part is this next section about when the input lacks key details. For a lot of prompts, context is important and I may not have given all the details necessary for it to respond with the best prompt. In those cases, it's supposed to realize that and ask for the needed details.
The next part covers when to include asking for different formats like tables, bullet points, or JSON. Then there's comparative analysis, multi-step explanations, creative brainstorming. This part was important.
If the user request involves idea generation or innovation, apply an appropriate thinking method, but do not force a rigid framework. So, I use the example of coming up with an invention or innovation. There's different frameworks people use.
One is scamper, substitute, combine, adapt, modify, put to another use, eliminate, rearrange, but there's multiple other approaches and frameworks. So, I wanted it to think of the best framework to use in any given situation when it applies. This is one of those areas where if you're making this for a custom use case, you can ask it to use the specific framework that's best for that.
Then step two is apply industry standard prompt engineering techniques only when needed. Most effective prompt engineering techniques are built into these models. Now, this is just a reminder to use them in the appropriate situations.
Things like fshot prompting, chain of thought reasoning, self-consistency prompting, role-based prompting, structured output formatting. Again, just reminding it to only use these when needed. They won't apply to every prompt.
And then step three is deliver the final refined prompt. There's a couple options here. If the prompt is already clear, simply refine it and provide the optimized version.
If additional information is needed, ask only the most necessary questions before refining the prompt. If the request is overly broad but usable, provide a refined version and mention that more details could improve accuracy. So that is the full prompt.
There's a lot in there, but it does work pretty well for general purposes. So I'll send that. How about explain coen economics?
I have no idea what that is, but I saw someone mention it on Twitter today. So that's the refined prompt with a lot more. That seems pretty solid.
Let's see how that does. There we go. Yeah, gave a pretty thorough answer.
And a good way to compare these side by side is you can actually edit the prompt. Then I'll write the more basic one. Explain Koji and economics.
And you can see them right next to each other. Yeah, it's just completely different answers. I'll try out another example.
What are the most common programming languages? Then it refineses a prompt and asks it to respond with a markdown table. Then there's how that answer would look.
How about one more? Give me app ideas I can vibe code. Yeah, that one's great.
Give me a list of fun and creative app ideas that are simple enough to vibe code. Quick to prototype, minimal setup, and fun to build solo. Focus on small scope, original or quirky concepts, and room for a playful UI or interactions.
Present that in a bulleted list with a short one to two sentence description for each. Yeah, that is perfect. Those are some solid ideas.
A bad advice generator. A hype mirror. One second sketch.
That's pretty fun. Single mysterious button that does something new every day. A tomaga tweet that only stays alive by tweeting or getting engagement.
Tiny compliment machine. All right, there's some good ones in here. Definitely did a good job with that one.
So, that works great and it's really good to have, but most people will have a couple very specific use cases they prompt regularly for. In those situations, it's best to make a specialized version of this rather than a general purpose one. One thing that I use chat GPT for a lot is creative brainstorming.
And even that is too broad a topic. So an even more specific example is scripting short form content. So I'll jump into the instructions first, then show it in practice.
This project will be used step by step. When I paste an article, tweet or concept, you will respond with step two. I will select a hook, then you will respond with step three, and so on.
We go step by step as a back and forth. And I have a bunch of steps all listed out in depth here. Step one is just the initial input.
Then step number two is hook ideas. There's 16 different categories like shocking statistic, intriguing question, bold statement, counterintuitive fact, and then 12 more. And over under project files, I uploaded a file that had three examples for each of these 16 hook types.
I created this a long time ago and could not find the original document anymore. And it doesn't let me open it up in here. I have no idea where that went, but that's what this document was.
But back to the instructions. Once I select a hook, it will give me visual hook ideas. Then it will give me key bullet points for the script.
The next step is to give ideas for curiosity gaps, unexpected twists, escalating reveals. This is all just to help with brainstorming. It will have a big list of ideas in different directions.
Some of the time I won't use any of them, but it will give me an idea. And then the next part is the rough script outline. Then some closing statement ideas like a call to action, loop creation, thoughtprovoking question, cliffhanger, and then there's a final review.
So that's a quick breakdown of how this works. Now I'll demonstrate it. How about we do the topic of this video, the secret to endless prompts.
How to use Chat GPT to enhance any prompt idea into the most effective prompt. And so this is the hook ideas, right? Shocking statistic.
95% of people are using Chat GPT wrong and missing out on its most powerful feature. What if your boring AI prompt could become a viral masterpiece with one tweak? You don't need creativity.
You need this Chat GPT trick. The best Chat GPT prompts don't come from AI experts. They come from beginners who know this secret.
What if one sentence could unlock an infinite stream of high performing AI prompts? Think of your prompt like a seed. Chadip is the one's not very good.
I used to spend hours stuck on a single idea until I discovered the trick that turned Chad GBD into my personal creative director. All right, so it keeps going. There's 16 of those.
Let's go with the counterintuitive fact number four. So that gives me the hook. Then it also has a refined hook.
So it's just another option with the same idea. With short form, you want to have a visual hook as well. They want to grab attention with what you're saying as well as what they're seeing on the screen.
So, it has a couple ideas for those. Let's go with number one. So, it gives me some key points for the script.
The problem with overengineering. Many people assume better prompts equals more complexity. Experts tend to overload their prompts trying to sound smart.
Then key point two, what beginners accidentally get right. Honestly, that sounds really engaging. Then key point three, the one trick that changes everything.
We'll say that sounds good and move on to step five. And there's a big list of things I could include. Some curiosity gaps.
Most people think they know how to prompt chatbt, but they're missing this. Unexpected twists. Escalating reveals.
There's relatable analogies. This one's pretty good. It's like talking to a genie.
If you're too vague or too specific, it backfires. That's really good, actually. Challenging assumptions.
Most tutorials over complicate prompting. Future implications. If you can master this one trick, you'll never run out of ideas again.
Cliffhers for various sections. Like, and here's the one tweak that changes everything. An interactive element like an on-creen poll.
say I like the second curiosity gap, the genie analogy, and the second [Music] cliffhanger. I'll just say that looks good and move on. There's some final ideas to wrap it up with.
Let's do number one. Then I have the full script outline with some different wording I can use. So, I never have it write the entire script for me.
I like to write it in my own voice, but it gives a bunch of bullet points and ideas for me to work from. Now, I wanted to walk through every step of that so you can see how effective and time-saving these can be. Like, there was some amazing ideas in there.
I never would have thought of myself. Like the comparison to a genie. I still really like that one.
So figure out what it is you do that is the most timeconuming or repetitive and see if you can work out a system like this for yourself. Or use chatbt to help you create that system. Use that prompt generator and ask it for a prompt to help you create it.
So you can save a ton of time for yourself this way. And if you want to go way more in depth on 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. You can get started for free using the link in the description or check out this video with 13 AI tools that can save you a,000 hours in 2025.
