The process of podcast synopsis generation traditionally has required creators, producers, and/or writers to understand the content, and manually write the synopsis, summarizing the salient points and highlights of an episode. This approach is time-consuming and may involve brainstorming sessions that discuss the main themes, and refine the text until it effectively conveys the podcast's message and tone (e.g., a comedic podcast may feature humorous language or hyperbole in its summaries).
This repository illustrates how to automate a large part of this manual and time-consuming process using the portfolio of services provided by Azure Cognitive Services, specifically Azure Speech Service and Azure OpenAI are used for transcribing and generating the synopsis, taglines, SEO keywords and translation into multiple languages. Incorporating AI to automate this process does not eliminate the role of human creativity or the importance of human involvement. Instead, it enables a significant acceleration in time-to-market by harnessing the power of AI. The final validation and approval of content remains the responsibility of human specialists before publishing.
This repository implements a high-level architecture that will take the audio of a podcast (steps 1-3), transcribe the audio into text using Azure Speech Services (step 4), and generate a synopsis based on the transcription using Azure OpenAI GPT-3.5 (step 5). Note that the version of GPT model used can be easily upgraded to, for example GPT-4, or future versions.
It consists of a React web app that allows a user to upload an audio file (podcast). The app uses an Azure function (WebApiHttpTrigger) endpoint to upload the audio file to a storage account, i.e. the raw-files container.
Azure function (AudioFileUploadedBlobTrigger) get triggered when a new file is added to the raw-files container. The function downloads the audio file and sends it to Azure Speech Services and transcribes the audio to text. It then saves the transcription to the same storage account, i.e. the transcription container.
The last Azure function (TranscriptionFileUploadedBlobTrigger), is triggered when new files are added to the transcriptions container. The function downloads the transcription file and sends it to four different Azure OpenAI methods that perform the following functions:
- Generate synoposis
- Generate tag lines
- Generate SEO keywords
- Generate synopsis translation in Mandarin Chinese, Brazilian Portuguese, German, and Latin American Spanish.
It saves the results as separate text files in the storage account open-ai-results container.
- Blog post with demo video for this solution.
- Prompt design examples for this use case.
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
To contribute, please start by creating a self-assigned issue giving a high-level overview of what you'd like to do. Once any discussion there concludes, follow up with a PR.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.