Using AI to write or brainstorm a book is not for everyone. However, for those who are curious, or those who struggle with productivity due to a disability, neurodivergence, burnout, and more, AI can often be a strong help.
- What the OpenAI Playground is
- What sets it apart from ChatGPT
- How to write a book using OpenAI Playground
- Other resources to use
Table of contents
- What is the OpenAI Playground?
- Why Use OpenAI Playground Instead of ChatGPT?
- How to Write a Book in OpenAI Playground
- Other Resources
What is the OpenAI Playground?
Unless you’ve been living under a rock, you probably have heard of ChatGPT and it’s parent company, OpenAI. But not a lot of people know what OpenAI’s Playground is.
OpenAI Playground is a tool that allows you to interact in more direct ways with models like GPT-3, GPT-4, and so on.
It allows people to experiment freely, get real-time responses, all while having a little bit tighter control over the output then what you could get with ChatGPT.
How Much Does OpenAI Playground Cost?
OpenAI Playground has a pay-as-you-go model. The pricing details vary depending on which model you use, but it is relatively unexpected. Here are the pricing models for GPT 3.5 Turbo, and GPT 4, which are the two you are most likely to use.
GPT 3.5 Turbo:
- Input (4K context window): $0.0015/1K tokens
- Input (16K context window): $0.003/1K tokens
- Output (4K context window): $0.002/1K tokens
- Output (16K context window): $0.004/1K tokens
- Input (8K context window): $0.03/1K tokens
- Input (32K context window): $0.06/1K tokens
- Output (8K context window): $0.06/1K tokens
- Output (32K context window): $0.012/1K tokens
There are a couple of terms that you should understand here:
- Input is referring to how much it costs for the AI to process what you give it, i.e. your instructions. These are typically half the costs of the output
- Output refers to what the AI generates
- Tokens is a method of measuring each specific part of a word. 1K tokens typically results in 750 words
- Context Window is the number of tokens that the AI can “remember” at a given time. The wider the context window, the farther back it can remember, but also the more it costs.
So all in all, OpenAI is not expensive. In fact, if you’re using a standard GPT-4 at 8K context, there’s a good chance that each chat exchange between you and the language model will only cost a few pennies.
And that’s using one of the more expensive models.
That said, many people don’t like the pay-as-you-go model, preferring the flexibility of a subscription model like what ChatGPT has, which gives you essentially unlimited words, but at a monthly cost.
Why Use OpenAI Playground Instead of ChatGPT?
But all this begs the question: why not just use ChatGPT? After all, it’s the same company, the same language models.
And that’s a good question.
But there are a few advantages for why you might want to use OpenAI Playground instead of ChatGPT. For example…
If You Don’t Use AI Often
If you’re not a regular user of AI, and you only dig into it occasionally, OpenAI Playground might be a better pricing model for you.
This is because if you’re only using the AI on occasion, the pay-as-you-go pricing model is unlikely to exceed the $20/month cost for ChatGPT Plus.
The System Box
OpenAI has a box that lets you input information you want the AI to remember. ChatGPT has something similar called “Custom Instructions”.
However, ChatGPT’s custom instructions has a character limit, whereas the System Box in OpenAI Playground has no limit (although bear in mind that this is a pay-as-you-go model and therefore the more you include in the System Box the more you will increase your costs).
But costs aside, this can be very useful when you have a lengthy style prompt, or a lot more context that you want the AI to understand.
It’s not a perfect solution, and ChatGPT features like Code Interpreter and Custom Instructions are lessening the need for OpenAI Playground’s features like these, but overall this System Box is a big reason why you might want to use Playground instead.
OpenAI gives you a little more control over how the large language model behaves. It includes more fine control over the following:
- The Model: this lets you indicate which of OpenAI’s models (3.5, 4, etc.) you can use.
- Temperature: indicates randomness. Sometimes more random can mean more creative but less predictable.
- Maximum length: let’s you set how much text can exist in the window (keep in mind that longer lengths will mean longer input text, i.e. more costly.
- Stop sequences: let’s you tell it when to stop. For example, when creating a list, give it a number, and the AI won’t go beyond that number.
There are other features as well, none of which are particularly relevant to authors with no technical background. But I’ve found that the model, the temperature, and the maximum length are frequently useful to adjust.
How to Write a Book in OpenAI Playground
My steps for how to write a book in OpenAI Playground are very similar to what you would do in ChatGPT.
The only major difference is that you might fiddle with the Temperature settings, or you might want to include a lot more information in the System box.
That said, here is my 7-step system for writing with AI. As always, you can use the AI for just one of these steps, or for all of them. Just pick where you have the most pain associated with the step, and use it for that.
For instance, I love doing the outlining, so I don’t use AI for that. But I’ve found it to be a great help in other areas.
Step 1: Brainstorm
Brainstorming is the creative groundwork for your novel. Here, you'll gather ideas, themes, and concepts that resonate with your story.
It's all about letting your imagination run wild and capturing the sparks of creativity.
Personally, I always start by brainstorming my premise and my ending, if I don’t know what they should be already.
Here are some prompt I’ve used:
Give me [NUMBER] high-concept pitches for a bestselling [GENRE] story with a unique twist, intriguing characters, and gripping emotional stakes.
Given the following premise and story information, give me [NUMBER] possible endings to this [GENRE] novel.
Step 2: Synopsis
The synopsis is a concise summary that encapsulates the main plot points of your novel. It helps you understand the flow of your story and ensures that the critical elements align cohesively.
You can expand upon your brainstormed premise and ending using a prompt like the following:
Given the following premise and story information, give me a highly detailed synopsis for a [GENRE] story in the traditional three act structure. Each act should be clearly labeled and should build toward the ending I've described.
Step 3: Characters and Worldbuilding
Characters are the heart of your story, and worldbuilding sets the stage on which they perform. There are many prompts you could use here, but here is a sample prompt for developing a character profile:
Write a character profile about the protagonist/antagonist/mentor of this novel:
Here is what we know so far about this character: [INSERT CURRENT SUMMARY OF CHARACTER]
Include the following elements: [INSERT DESIRED ELEMENTS HERE]
Step 4: Outline
Outlining offers a structured view of your novel's framework. It can be as detailed or as broad as you need it to be, providing a plan that you can follow as you write.
From your synopsis, you can expand it further into a fully fleshed out outline:
Using the following synopsis, create a detailed summary of the story, fleshing out additional details, and breaking it into parts using the [OUTLINE METHOD OF CHOICE]: [INSERT SYNOPSIS HERE]
Step 5: Story Beats
When creating a story with AI, it helps to have an intermediate step between the outline and the actual writing. So fleshing out the outline into individual story beats (essentially a highly-detailed outline), can look something like this:
Take the following chapter summary, and generate a list of 12 highly detailed action beats for a script with additional STORY INFORMATION to fully flesh out the chapter. Make sure to always use proper nouns instead of pronouns.
[INSERT CHAPTER SUMMARY HERE]
Step 6: Style
A style prompt is possibly the most important prompt on this list, as it will affect how the prose sounds when you use AI to generate prose.
Style prompts can be lengthy, depending on how you use them. I've seen some that go on for hundreds of words, but here is one example of a shorter one thta could work:
First person past point of view of [MAIN CHARACTER], show don't tell, deep point of view. Realistic dialogue. Stronger verbs. Lots of conflict, drama, and description. Avoid mushy descriptions/dialogue.
Step 7: Prose
Last but not least, it's time to create the prose prompt. Now many, if not most, authors will not be using AI for this step, and that's okay. In fact, I would say that AI is best used as a brainstorming tool.
However, there are many who DO want to write prose with AI, and I would do it something like this, using 2-3 story beats at a time.
Write 600 words of a chapter using the following details:
Genre: [ENTER GENRE HERE]
Tone: [ENTER TONE HERE]
Point of View: [ENTER POV/TENSE HERE]
Setting: [ENTER SETTING HERE]
Key Characters in This Scene: [ENTER CHARACTER DETAILS HERE]
Style: [ENTER STYLE HERE]
Story Beats to Cover: [INSERT STORY BEATS]
AI is an ever-shifting resource, and so there are a few additional resources that I would recommend if you want to stay on top of it.
The first, is my own YouTube channel, The Nerdy Novelist (a personal project, not associated with Kindlepreneur) where I provide almost daily videos about the goings on in AI.
The second is Future Fiction Academy, which is a paid subscription that gives you access to daily labs that are constantly exploring and keeping subscribers up to date on what AI can do.
Plus, it gives you a TON of instruction if you’re just starting out and just want to learn the basics.