How I Work With ChatGPT and Plugins: The Jack Righteous Creator Workflow

Free creator training · October 2026 · A practical starting point for independent creators and collaborators

You made something with AI. Now make it useful, clear, and worth building around.

When somebody asks me how to get better with ChatGPT, I don't start by showing them every setting, every model or every Plugin I use. I want to know what they are trying to create. A song? A book? A video? A creator business? And I want to know what they have already tried.

That matters because the problem is rarely a shortage of AI tools. More often, an idea goes into ChatGPT, a convincing answer comes back, and we move too quickly. We skip the questions that make the work better. Then we spend money generating the wrong thing, or we end up with files we can't explain or use.

I've learned that lesson working with AI video. You can give a generator a carefully prepared image and still discover that it cannot reliably reproduce the words you put inside it. By the time you realize the problem, you've used credits. That is not just a prompting issue. It's a planning and review issue.

I've also built reusable AI specialists for work I do repeatedly. They help me follow a process more consistently. But those specialists only became worthwhile after I understood the work they needed to support.

That's what I want to teach you here. Not how to become a copy of me, but how to develop your own judgment and work with a repeatable method.

What we'll make together

We'll follow one example all the way through: a 30-second vertical video promoting a book. You can use the same method for a song teaser, a creator introduction, a short story, an online product, or another project you have permission to develop.

By the end, you should have a clear creative brief, a storyboard, one carefully chosen AI-generation test (or a ready-to-test prompt if you don't want to spend money), an honest review, and a record someone else could use to continue the work.

You don't need to buy video credits or install a Plugin to start. If you're already comfortable with ChatGPT, move directly into the worked example. If you're new, the free ChatGPT beginner guide explains the basic controls in more detail.

Part 1 — Find the work worth doing

Imagine an independent author has finished a book about finding a new direction after a difficult period in life. That's a fictional training example, not a description of anyone's unpublished manuscript.

The first request might be: “Make me an amazing trailer for my book.”

ChatGPT will give you something. It may even sound convincing. But it cannot know which story elements are approved, how the author wants to be represented, or how much the production can cost. Starting here asks the software to make decisions that belong to the creator.

I would begin with a short conversation instead:

I'm preparing a 30-second vertical teaser for a book. The intended readers are adults interested in personal stories of perseverance. I want the video to create curiosity and feel hopeful without becoming dramatic or sentimental. I will provide the approved summary. Don't invent a scene, quotation, title or endorsement. Ask me only for information that would change the creative direction. First help me choose one message worth communicating.

Notice what's different. We're not telling ChatGPT to produce the video. We're asking it to help us make the first creative decision.

For your own project, write down what you're making, who it's for, what you want them to feel or do, and what must not change. Four clear answers are more valuable than a page of vague adjectives. If you don't have approved source material yet, stop at a concept brief. Do not pretend you have a finished story to promote.

Your first checkpoint: Can another person read your brief and tell what the piece is supposed to accomplish? If not, improve the direction before going further.

Challenge the first answer

Now ask ChatGPT for two possible ways to communicate that message in 30 seconds. Make it explain the trade-offs. One approach could rely on a narrator and restrained visuals. Another might use an on-screen character and more movement.

I would favor the simpler approach for the first test, not because simple is always better, but because it gives us fewer variables to pay for and diagnose. If we test a complex scene with several faces, moving objects, lip-sync and visible book text, we may not know what caused the failure.

The next instruction is the one I don't want you to skip:

Challenge the proposed approach. What did you assume about the book, the video generator or the audience? Which details have not been verified? What is the cheapest useful test? Give me one reason to reject your own recommendation.

That's the difference between using ChatGPT as a question box and using it as a creative assistant. It should help you see weak points, not just tell you that every idea is great.

What good work looks like: You can explain why you chose the direction, what you rejected, and which parts still need your judgment or an official source check.

If this becomes an ongoing project, create a ChatGPT Project and keep the approved brief, permitted reference material and decisions together. Projects are useful for continuing context; they aren't a substitute for verifying facts. The JR guide to using training inside a ChatGPT Project goes further, but it may require eligible member access.

Part 2 — Plan before you spend

This is where prompting starts to cost money if we aren't careful. I want ChatGPT to help us prepare the production before the expensive video generator gets involved.

Here is an illustrative first storyboard for our fictional book. It is deliberately modest and must be adapted to the real approved synopsis before production.

This is not a trailer ready to publish. It's a production hypothesis. We still need the author's approved source material, the actual book title, appropriate permissions, the intended destination, and a voice that fits.

Here's the kind of ChatGPT request I'd use before deciding which video app to open:

Using the approved brief and the four-shot outline, prepare one six-second test shot. Keep the prompt appropriate for a vertical 9:16 video. Describe the subject, framing, movement, light and mood. Avoid readable words, logos, dialogue, additional people and effects that are not needed. Identify what could fail and how to evaluate the test. Do not claim to know the settings of a specific app until we have checked its current instructions.

And here's what a first tool-facing test prompt might look like, once the app and its settings have been verified:

Vertical 9:16, approximately six seconds. A closed unmarked notebook rests on a natural wooden table beside a window. Warm morning light. One slow, stable camera push toward the notebook. Calm, cinematic realism. No people, logos, captions, readable writing or sudden camera motion.

Why test this shot first? It asks for one subject, one movement and one lighting condition. We can inspect framing, stability and mood before asking for a harder scene. If a six-second shot doesn't follow the direction, we should learn from it before spending credits on the whole video.

Your own version can be more creative. It should also be testable. A good prompt doesn't guarantee success; it makes success and failure easier to recognize.

Spend credits like a producer

Here's where I want you to think like a producer, not just somebody enjoying new software.

Before a paid render, confirm the current account allowance, pricing, supported duration and aspect ratio, source upload permissions, and commercial terms. If the tool is new to you, use its own current documentation. ChatGPT can help research these facts; it should not invent them.

Then set a clear limit. Decide what you are testing, what counts as a usable result and how many attempts you're willing to fund. For our notebook shot, the first pass might need a stable camera, no unwanted text and a mood that matches the brief. If the movement fails, change the movement instruction. Don't rewrite the whole campaign or generate five unrelated versions.

Another important decision: separate what the video model should create from what an editor should finish. A generator may be useful for motion and atmosphere. An editor is often the safer place for exact titles, credits, captions, URLs and final timing. Likewise, narration should be reviewed for pacing and permissions rather than assuming the generated voice will read a full script naturally.

If you're not ready to spend anything, you can complete this lesson with a storyboard, a script and a test plan. That is a legitimate result. You have learned to prepare the work instead of paying to find out what you forgot.

Your production checkpoint: What specific result would make you approve the next render? If you can't answer, you aren't ready to spend more credits.

Part 3 — Make the useful part repeatable

Notice what we've done so far. We haven't needed a Plugin. We've needed a clear idea, a disciplined conversation and a controlled test.

Now imagine you prepare five videos. Every time, you ask ChatGPT to check source accuracy, audience fit, shot timing, unwanted text, permission risks and likely credit waste. That repeatable review is a good candidate for a Plugin. The creative concept itself is not something I would want a Plugin to decide for you.

Here's the distinction I use:

• A Chat helps with the question or decision in front of you.
• A Project keeps the approved context and decisions for work that continues.
• A Plugin gives a repeated job a consistent method and, when appropriate, access to supported connected apps.
• ChatGPT Work helps carry out larger, supported multi-step tasks. Actions that affect real accounts still need suitable access, approval and verification.

You don't have to use every layer. If you only need a storyboard, stay in Chat. If you're working on a book for several weeks, a Project may help. If you're checking every storyboard against the same criteria, a specialist becomes worthwhile.

If an appropriate Plugin is available in your ChatGPT account, ask it to review, not render:

Review this storyboard against my approved brief. Identify the three most important production risks, explain the evidence for each, recommend one practical correction and identify which decisions still require me. Do not buy credits, publish, edit my accounts or imply you tested a video you haven't seen.

If no suitable Plugin is available, use exactly the same request in ChatGPT. You are still learning the method. Private JR Plugins are not automatically included with an article or membership, and I would never advise sharing private account connections just to complete this exercise.

For a deeper look at reusable specialists, see Custom GPTs for Creators: From Repeated Prompts to Plugins.

Build your first private specialist

When you're ready to make a Plugin of your own, keep the first version small. I would call ours Video Brief Reviewer. It has one job: decide whether a proposed short-form video brief is ready for a first production test.

It needs an input (the approved creator brief and storyboard), a method (check message, source facts, timing, feasibility, rights and budget), an output (ready, revise or hold, with reasons), and boundaries (no spending, posting or accessing private material without authorization).

That specification is more important than which menu you click to create it. Where available in your account, use the Plugin Creator tool to build a private specialist from these instructions. You can also save them as a reusable text template. OpenAI is transitioning from custom GPTs to Plugins, and the migration timeline, sharing and permissions vary by account. See the official migration guidance before planning around older GPT creation.

Don't call it ready just because it gives a polished response. Give it a good brief and a deliberately weak one. Does it catch a fabricated claim? Does it flag missing rights approval? Does it stop if asked to publish without permission? Does it preserve the creator's intentions instead of rewriting everything into the same generic style?

Now you're testing a system, not admiring an answer. Keep your approved instructions and test examples outside the Plugin so you can maintain or move the method later.

The optional Creator Pro specialist-building lesson covers more advanced implementation for eligible members. It isn't required for this free first exercise.

Part 4 — Finish, hand off and improve

One of the most valuable habits in my workflow is not leaving the result buried in a chat. When a creator or collaborator hands off a project, I want to see the evidence—not just hear that AI said it was ready.

For your first assignment, prepare a simple record:

Project: What you set out to make, and for whom.
Approved direction: The brief and source material you used.
Current asset: The storyboard, test prompt or reviewed render, with a version or link.
What worked: A specific observation, not a claim of perfection.
What failed or changed: The correction you made and why.
Cost and rights: Credits actually used, relevant permissions and anything unresolved.
Decision: Draft, for review, approved or released—and who made that decision.
Next action: One useful move, not a new list of twenty tasks.

If you're collaborating with me or someone else, this handoff lets us review the work and improve the process. It also provides the evidence we need before turning your method into something reusable. Never include passwords, unrelated private conversations or another creator's protected source material without permission.

Your first JR assignment

Start with one creative idea. Complete the brief, the four-shot storyboard and one test-shot prompt. Then challenge the plan and document one revision. You can finish without generating any paid video. If you do render a test, save the actual result and what it cost.

When you can explain why your revised approach is better, you're ready to consider a Project or a reusable specialist. When you can show the same review method works on different projects, you're ready to build a Plugin around it.

That's the direction I want you to take: Make something. Understand what worked. Improve it. Then build a system around the part worth repeating.

Create What You Love | Love What You Create.

Last reviewed October 26, 2026. Platform features, account eligibility, tool credit policies and legal permissions can change; check current provider terms before spending, sharing or publishing.