What level of GPU is required to generate subtitles for a 10-minute video within 3 minutes

[kewer127]

I use whisper to generate video subtitles. However, even with my decent Linux server (AMD Ryzen Threadripper PRO 5965WX s (48) @ 3.800GHz), it takes 30 minutes to generate the text using the default medium model with all CPU parallelly. I have no concept of GPUs. What level of GPU would I need to generate subtitles for a 10-minute video within 3 minutes?

[phineas-pta]

just buy a recent nvidia gpu that can fit the whisper model u using

or try whisper.cpp

[kewer127]

Thanks! whisper.cpp is way faster on my server! Whoa!