Support for Audio Transcription in openai_dart Module

[Stefano-Trinca]

trafficstars

Feature request

Add functionality to the openai_dart module within the LangChain library to support audio transcription. This feature would allow users to transcribe audio files using OpenAI's transcription models directly through the openai_dart package.

Motivation

Currently, the openai_dart module does not provide support for audio transcription, which limits its usability for applications that require converting audio to text. Adding this feature would enhance the module's capabilities and make it more versatile, allowing developers to utilize OpenAI's powerful transcription models in their Dart applications. This would be particularly useful for creating applications that need to process spoken language, such as voice assistants, automated note-taking tools, and accessibility features for the hearing impaired.

Your contribution

I can help by testing the new feature once it is implemented and providing feedback.

[davidmigloz]

Hey @Stefano-Trinca,

Thanks for opening the issue. This is currently blocked by this: https://github.com/davidmigloz/langchain_dart/issues/403#issuecomment-2139146023

[jbienzss]

Though we don't have a way to upload a file, I'm still very interested in using Streaming-Transcription as documented here:

https://platform.openai.com/docs/guides/realtime-transcription

Is this possible with the current library? I noticed this recent PR by @popcorn-cpu:

https://github.com/davidmigloz/langchain_dart/pull/698

But am I correct in understanding that's still going against Realtime-Conversation rather than Realtime-Transcription?

My main reason for wanting streaming transcription is that I want to run a tool on every sentence a user speaks (not just every turn). If there's another way to do this with Conversation-Client, I'd love to know that too. Thank you!