whisper image is broken on main

[axel7083]

When testing containers/podman-desktop-extension-ai-lab#1512 I got the following error when running the curl command:

ghcr.io/containers/whispercpp:latest

$: curl -X POST "http://localhost:38285/inference" -F "file=@./004NTWY_U1_RC.mp3"
{"error":"FFmpeg conversion failed."}

whisper image from v1.1.3

$: curl -X POST "http://localhost:38285/inference" -F "file=@./004NTWY_U1_RC.mp3"
{"text":" Unit 1, page 14. Real conversations.\n Hello. Hi. What's your name?\n Claudia, what's your name?

[axel7083]

Testing the main branch for the audio to text recipe

Server logs

[in#0 @ 0x7f1a54d112c0] Error opening input: No such file or directory
Error opening input file whisper_server_temp_file.wav.
Error opening input files: No such file or directory

[axel7083]

Looking at the history of whisper model server I saw that d029eba introduced a changed in the group permissions introduced by #681

After checking the previous commit, and building the image, I can confirm that this PR introduced the regression.

cc @johnmcollier