← Video Library · Content & Video
Stop Wasting Recordings! Turn Transcripts into Content Gold
Published September 4, 2023 · 111 views on YouTube
Key takeaways
- A recorded transcript from a mastermind session, training video, group call, or one-on-one coaching call can be turned into a blog post and micro-content using a prompt recipe stack.
- The two inputs the recipe needs are a transcript and the recipe stack itself, with the Avatar, the Niche, and the context being the parts you actually customize per run.
- The recipe runs as a sequence of prompts (executive summary and key takeaways first, then a self-prompting rewrite pass, then micro content, then video scripts) that can also be run individually rather than all at once.
- Output quality depends on the quality of the context provided, so anything brand-specific, like social handles or a brand voice, should be added to the pre-prompt before running the recipe.
- The same recipe can be re-run with retry, or extended afterward with follow-up prompts, such as asking it to draft an email recap from the executive notes and blog post.
How do you turn a transcript into a blog post with Claude?
Feed a recorded transcript, such as one from a mastermind session, training video, group call, or coaching session, into a prompt recipe stack in Claude along with the Avatar, the Niche, and the context. Run the recipe starting with an executive summary and key takeaways prompt, then let it self-prompt through refinement, micro-content, and video-script stages, editing and re-running any step as needed.
What do you need before running the recipe?
Two things: a transcript and a recipe stack. The transcript can come from a mastermind session, a training video, a group call, a private one-on-one coaching session, or similar recorded content. The recipe stack is a set of prompts you keep and reuse, where the parts you actually change per run are the Avatar (who the content is for), the Niche, and the context (the transcript itself). In the video, the Avatar used is business professionals in the generative AI niche, with the context set to the transcript from a recent mastermind and machine announcement session.
How does the prompt recipe stack work?
- Paste the transcript into the context field of the recipe and set the Avatar and Niche, keeping the rest of the prompt the same.
- Run the recipe in Claude, starting with the first prompt, which produces the key takeaways and executive summary or notes.
- Let the recipe self-prompt into a rewrite pass that adds additional insights and can be reframed into a specific perspective (for example, rewritten in the creator’s own voice).
- Continue into the micro-content prompts, which can also be run as a separate step after the main content is finalized.
- Move into the video scripts stage, which recaps the keynotes, insights, and variations already generated from the transcript.
- Review the output and use retry to rerun any step if the format or quality isn’t right the first time; it often takes a few runs to get it dialed in.
Each prompt in the stack does not have to be run all at once in a single command-stacking sequence. Prompts can be run individually, edited, and modified as you go.
Why does the quality of the output depend on the transcript and context?
Because the recipe generates content directly from whatever is in the call, training session, webinar, or mastermind that was transcribed, the substance of that source material drives the substance of the output. Anything specific that should appear in the final content, such as social media handles or a defined brand voice, needs to be added into the pre-prompt so Claude has that framing before it starts writing, rather than being expected to infer it.
Can you extend the recipe beyond the blog post and micro content?
Yes. After the main recipe output is finalized, you can add further instructions, such as asking Claude to reference the executive notes and blog post to draft a complete email recapping what was covered on the call, or to rewrite the blog post afterward with more in-depth review and insights. The workflow shown uses Claude, but the same approach can be applied in ChatGPT using GPT-4.
Full video transcript
Hey, Darby here from Gen AI University, and in this video I’m going to be breaking down how to take a transcript and turn it into a blog post that is refined and also can be turned into micro content and short blog posts. Use this for mastermind sessions, training videos, group calls, private one-on-one coaching sessions, things of that nature.
The two things you really need are: one, a transcript. I’ll be selecting one from our recent mastermind and machine announcement for Gen AI University. And two, our recipe stack that we’re working on here. Keep in mind that the real things you want to provide within this recipe are the Avatar, the Niche, and the context. I’ll be using business professionals in the generative AI niche, and the prompt is going to remain the same, with the context being that of the transcript that I added in.
Once I’ve done that, I’m going to go ahead and select enter inside of my Claude interface. While it might not run perfect the first time, within two or three runs it does a pretty good job at executing it. I wanted to start with the key takeaways and the executive notes, and from there we’ll refine and rerun the recipe and edit it.
Once I’ve got a good run of that recipe against the content, what I’m looking for… it’s self-prompting itself at this point to rewrite it with additional insights that it’s taking away from the initial version of it. You can see here with this final prompt, prompt five, I had it rewrite it into the perspective of me. As you’re going through a recipe like this, you can edit and modify it, and you frankly don’t have to run all these prompts at once. You could run each of them individually, which is totally fine as well. This is just a demonstration for how to run them all in our command-stacking sequence.
Now we’re moving on to the micro content, and again this could easily be a separate prompt that after you’ve finalized the content above, you then add these different prompts and, one by one, just follow that same copy-and-paste workflow. It’s moving on to video scripts, and just to recap what we’ve covered: we’ve got these different keynotes, insights, variations of the content, and then a broken-down version of what was essentially covered inside of this transcript.
It’s important to also point out that the quality of the outputs here is coming from the context that we’re providing it, which is that of the announcement video, the training session, the webinar, the mastermind call. Whatever is happening inside of that call is obviously critical for the content you’re going to be getting out here. But also, if there’s anything specific, like social media handles, brand voice, things of that nature, you’d want to add that into the pre-prompt and frame the context for Claude so that it is articulating things that are on brand, if that information is necessary.
Just for example’s sake, say I didn’t like that output and I wanted to rerun it again, I can just simply go and select retry, and it reruns the content brief again, as you can see here. You’ll notice that sometimes the format is a little bit different, and again it’s why we continue to refine these recipes, to fine-tune them to produce the outputs that are most desired. But overall this is relatively in the right direction as to where we’re going to want to be going. Like I said, it might take a few runs to get it perfect.
If you want to copy it, you simply grab that content, example output, share right here. This is just to save your work. If I want to rerun it again, it’s as simple as just clicking retry. This, by the way, is the process that I used to work with Claude, refining. With ChatGPT, Claude tends to have longer, seamless outputs and can reference a boatload of context with the transcript and the recipe that we’re providing it.
Okay, the second output. Now what I want to do is I want to do one more time on this. So we started that prompt one with that executive summary and overview. Now the AI is providing suggestions based on that. We get to prompt five, where it’s coming in from my perspective. This would actually work for an email as well as a post or an announcement or anything of that nature, and then it moves into the micro content.
So I’m again running and re-running and testing, tweaking, refining as a recipe in general, and we will continue to build and grow over time. But all in all, taking an existing transcript and plugging this recipe in, you can see how you can very quickly get a bit of content from it that used to provide additional value to those who have participated, or need to consume content but maybe don’t have the minutes or an hour or however much time to attend the call live, or simply want to have a quick recap of what was covered. Then you take the different prompts inside of the recipe and use them at your own disposal, for your purpose, in the way you like, and feel free to remove and not necessarily use the additional prompts, like micro content, for example, if the goal is not to produce this over social media, things of that nature.
But now let’s just say we want to take things a step further, once we’re done with this output of the recipe. I’ll just say something simple like: reference the executive notes and blog post to craft a complete email from Darby to send out to the list with a recap of what was covered on this announcement call. Also, rewrite the blog post after the email with more in-depth review and insights from the video. Let me try that once. Thank you.
This is much better. So what I’ll do is I’ll take this email and modify it and edit it just a little bit before it goes out. Now, rewrite a blog post. There we go. Hope you found that useful, and following that workflow, it’s a simple Claude place to be applied, probably ChatGPT using GPT-4, and give it a shot yourself. Just remember that this is a recipe that you are more than welcome to modify, change, and enhance for your own desires and or your own specific needs, more so. Adding in that Avatar and Niche along with the context of the transcript is really critical for success in any of the content you create with AI. So until next time, keep calm and AI on.