I found whisper API perform better than colab run whisper large-v3

API
1

i do have examples but indeep i dont think i need submit.

shortly,

for medical videos,

  1. mobitz type 2 is quite correct in API,
    but as mobits in large-v3.

  2. actually besides this,
    a medical video of about 30mins, i split into short segments and merge them back.
    for api, everything works.
    for large-v3, certain part is NOT useful, only after human edition will make sense.

since i use the same python scripts to split and merge the files/srt,
i am sure large-v3/local run is inferior to the API.

did openAI claim that both should be exactly the same?
thanks

ps: i am not concerning the fee, i am concerning on the quality.