Speech2text_example does not work on HAILO-10H with HailoRT 5.2.0

Hello,
I am using HAILO-10H with HailoRT 5.2.0.
In this environment, the hailo-apps/gen_ai_apps/simple_whisper_chat example works fine.

However, after building the hailort_examples/speech2text_example, it does not run successfully.

I used the same HEF files and audio files as in simple_whisper_chat.
For the audio file, I tried both the original and a converted version using the following command, but neither worked:

sox audio.wav -r 16000 -c 1 e float -b 32 -L audio_16khz_mono_f32le_norm.wav norm

This means 16 kHz, mono, float32 little-endian, normalized to the range -1.0 to +1.0.

Could you please advise why speech2text_example is not working on HAILO-10H, and what steps I should take to make it run?

I tried on Ubuntu as your instructions helped me figure out where to get the Whisper HEF file from. Strangely for me, it runs, but can’t recognize the words:

$ ./speech2text_example ./Whisper-Base.hef ./audio.wav
Starting Speech2Text…
[0, 2] [Music]

I’ll try debugging it tomorrow.

Thank you for your reply.

There was a misunderstanding in my previous message. The application also runs correctly in my environment. However, just like in your case, it does not recognize any speech.

Worked fine for me on a raspberry pi 5 and a 10H. I had the same problem but found I was using the incorrect Whisper model.

I also was using HailoRT 5.2.0