I ran a sample program using this library based on Windows ARM. When processing a 13 minute source audio, it takes 11 minutes to complete, while the whisper official command-line program only takes less than two minutes. Is there any optimization method available? By the way, the test is based on the same ggml model.
I ran a sample program using this library based on Windows ARM. When processing a 13 minute source audio, it takes 11 minutes to complete, while the whisper official command-line program only takes less than two minutes. Is there any optimization method available? By the way, the test is based on the same ggml model.