From Speech to Blog Post: Testing OpenAI's Whisper APIA blog post by Jesse Luoto

Have you ever wished for a tool that could transcribe your spoken words into a beautifully crafted blog post? Well, OpenAI's Whisper API might just be the answer!

Today I created a tool called babble2blog that takes spoken text input and creates a beautiful blog post out of it. But, is it really that simple? Let's explore the process and potential challenges of using this technology.

Sometimes doing tools for yourself is more enjoyable than the task itself?

The Process

OpenAI's Whisper API takes spoken text and converts it into a text format that can be fed into their GPT-3 model. Once prompted, the model generates a full blog post from the text that was inputted. While this technology is impressive, there are a few challenges that need to be addressed.

The Challenges

One of the biggest challenges is losing the nuances of spoken language when it is transcribed into text. Intonation, repetition, and mannerisms are all lost in the process. Additionally, spoken language is not typically in a blog post format; it can be difficult to convert raw spoken language into a properly formatted and structured blog post. However, if successful, this tool could greatly decrease the time it takes to write a blog post.

The Verdict

Did it work? See for yourself! This blog post was created using the tool I created, with only minor finalizing edits (and the meme, obviously) to GPT-3's output.

Listen to the original audio recording of the blog post:

(Note the super bad quality – this was recorded with my Macbook's built-in mic while recovering from a cold)

You can also try out the tool yourself by visiting the Github repository:

https://github.com/jehna/babble2blog

Let me know what you think! Please tweet or leave an issue on Github if you have any feedback.

Tweet

Be the first to know from new blog posts

Subscribe to the mailing list to get priority access to new blog posts!