Thank you for this review and the benchmark at the bottom. It seems that Turbo gives the best bang for the buck at the moment. Someone pointed out that OpenAI still recommends Large-v2 in the OpenAI Platform documentation, and your benchmark further supports the view that Large-v2 is slightly more accurate than Turbo, but at a significant speed cost.
Based on my experience benchmarking and using the different versions, turbo is a really solid model while being notably cheaper and faster to run so I would recommend it as a starting point. To get more accurate transcripts, fine-tuning it greatly helps even when we train on less than 10 hours of audio.
Thank you for this review and the benchmark at the bottom. It seems that Turbo gives the best bang for the buck at the moment. Someone pointed out that OpenAI still recommends Large-v2 in the OpenAI Platform documentation, and your benchmark further supports the view that Large-v2 is slightly more accurate than Turbo, but at a significant speed cost.
https://platform.openai.com/docs/guides/speech-to-text/introduction
I'm glad you liked it!
Based on my experience benchmarking and using the different versions, turbo is a really solid model while being notably cheaper and faster to run so I would recommend it as a starting point. To get more accurate transcripts, fine-tuning it greatly helps even when we train on less than 10 hours of audio.