Yesterday, OpenAI announced public access to their ChatGPT and Whisper API endpoints and we couldn't be more excited. These tools have been greatly accelerating our team's workflows for the past few months. Gaining the ability to access them programmatically is a game changer.
However, we know that for some folks, getting started with these APIs might be more difficult. We want to ensure that anyone working data can easily access the power of these services. That's why today, we're adding two new Blueprints to the application: "ChatGPT - Generate Response" and "Whisper - Transcribe Audio with API".
ChatGPT - Generate Response
This general purpose Blueprint allows anyone to leverage the power of ChatGPT as a part of their data workflows. Provide any prompt you'd like and watch as the magic of ChatGPT generates responses. These results can be stored as files and used in downstream tasks.
You can use this to do things like:
- Write Emails
- Summarize datasets
- Generate SQL Queries
- Translate text between languages
- Create comments or tests for your code
Whisper - Transcribe Audio with API
While we had previously built out a Blueprint to help transcribe audio with Whisper (and even made a few videos about it), the new Whisper API endpoint makes transcription of audio even easier.
With Whisper, you can provide an audio file and quickly transcribe the entirety of the contents into the original language that the audio was spoken in. The speed is relatively quick too - with most transcriptions taking less than 1/5 of the time of the actual audio file. Plus, you can also translate any audio directly to English!
This transcription file can then be stored or uploaded to external services to use however you see fit.
Powering Up with Parameters
Through the use of Shipyard's webhook parameters, your use cases can start to really shine and become more dynamic.
Let's say that you build out a workflow to download an audio file, transcribe it with Whisper, translate it with ChatGPT, then upload the results to S3.
Using our webhook parameters, you can use this same workflow again and again, each time passing a unique audio file to be transcribed. This gives you one central location to transcribe all the files for your business, tracking the performance of each unique file that gets processed.
This is only the beginning of our deeper integrations with OpenAI's APIs. We believe that future of data work will be accelerated by mixing and matching AI directly with your own data.
Over the course of the next few weeks, we'll be working to build out more specific use cases with ChatGPT that require less initial prompting on your end. The end result will be faster workflows for your business, powered by AI. Have a need? Let us know!
Get Started Today
ChatGPT and Whisper Blueprints are now available to all users and can be tested with any account. Sign up for our free Developer Plan to start automating your data workflows with more AI tools.
We're looking forward to seeing how users will take advantage of this functionality to quickly launch, monitor, and share data pipelines!