Transcript and translate audio files automated flow

After a few days of battling small but annoying technical issues (you know the ones!) and carefully choosing the best infrastructure, I finally have my step flow ready.

One thing I’ve learned during this process is that my developer background often leads me to think of large, complex solutions—which, while exciting, aren’t always necessary to deliver the best results at this stage. Keeping it simple but effective has been the key!

Here’s the current workflow:

1️⃣ File Uploads: Files are uploaded to a Google Drive folder.

2️⃣ Format Conversion: Files are converted to WAV format using Zamzar.

3️⃣ Transcription: Both Whisper and Google Speech-to-Text API process the files.

4️⃣ Quality Check: ChatGPT compares the two transcriptions and selects the most accurate one.

5️⃣ Translation: The transcription is passed to both Deepl and Google Translate API.

6️⃣ Final Review: GPT-4 evaluates the translations, checks the original transcription, and produces the most accurate translation.

The result? A smart and efficient workflow that leverages the best tools to optimize quality and results. 🌟

🚀That said, there’s one use case I still need to address: files larger than 25MB. These need to be split into smaller chunks before following the same flow.

It’s a challenge I’m excited to tackle next!

I’ll keep you posted with the outcomes and new developments as this project evolves. 🎉

#workflowautomation #ai #gpt4 #googlecloud #deepl #whisper #optimization

3 comments