90% savings using optimised Whisper API and distributed computing (qblocks.cloud)

As a developer, I achieved 90% cost-effective transcription and translation by Optimizing OpenAI Whisper and using an innovative approach to get GPUs.

TL;DR :

1.CTranslate2 is used to optimize the OpenAI Whisper model for efficient inference with Transformer models.

2. It can be easily installed on a Q Blocks GPU instance with a simple command.

3. The optimized Whisper model can be run on a Q Blocks decentralized 3090 GPU instance.

   A comparison showed that using the optimized model on Q Blocks resulted in a 12x cost reduction compared to the default model on AWS.