How to successfully transcribe audio files using Whisper for OpenAI in Python?

Question

I´m trying to use Whisper model but occours an error saying:

UserWarning: FP16 is not supported on CPU; using FP32 instead warnings.warn("FP16 is not supported on CPU; using FP32 instead") file not found. exiting process.

my code:

import whisper
import os

# created this logic to get the real path to the file

current_dir = os.path.dirname(os.path.abspath(__file__))
file_name = "audio.wav"
file_path = os.path.join(current_dir, file_name)
print(file_path)

# the transcript logic:

listen = file_path

def transcribe():
    model = whisper.load_model("large")
    result = model.transcribe(listen)
    print(result["text"])
    return result

try:
    result = transcribe()
    with open("transcript.txt", "w") as file:
        file.write(result["text"])
    print("file saved")
except FileNotFoundError:
    print("file not found.")
except PermissionError:
    print("Not permitted to save file.")
except Exception as e:
    print("An unknow error occured:", str(e))
finally:
    print("exiting process.")

Thanks for now

By the way, I tried this:

current_dir = os.path.dirname(os.path.abspath(__file__))
file_name = "audio.wav"
file_path = os.path.join(current_dir, file_name)
print(file_path)

and the path by itself = "C:\Users\rene.pessoto\Desktop\Audio\audio.wav" and ("C:\Users\rene.pessoto\Desktop\Audio", "audio.wav") but I can´t finish it yet.

*the file has 39 minutes and its size is 214MB

This question is similar to: whisper AI error : FP16 is not supported on CPU; using FP32 instead. If you believe it’s different, please edit the question, make it clear how it’s different and/or how the answers on that question are not helpful for your problem. — Skopyk
– Skopyk, Commented Jan 17 at 16:18

jitheshKuyyalil · Accepted Answer · 2024-08-09 04:44:10Z

0

I get the same warning except the "file not found. exiting" part. But the code runs smoothly and transcription works fine. The python openai-whisper module is still working without any API, locally. We need to just install openai-whisper using pip. So there is no question about change of API.

We need to look at the output of print(file_path).

current_dir = os.path.dirname(os.path.abspath(__file__))

is only helpful if your source file is in the same folder as your audio file. If not your file wouldn't found and you'll get 'file not found' error.

answered Aug 9, 2024 at 4:44

jitheshKuyyalil

615 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Skopyk · Accepted Answer · 2025-01-17 16:18:09Z

0

The code seems fine. The first and foremost problem is that your machine might lack either software or hardware necessary for whisper to work. See this whisper AI error : FP16 is not supported on CPU; using FP32 instead

answered Jan 17 at 16:18

Skopyk

438 bronze badges

Comments

mb7 · Accepted Answer · 2023-06-27 12:18:54Z

-1

Instead of using

result = model.transcribe(listen)

try to use this

options = whisper.DecodingOptions(fp16 = False)
result = whisper.decode(model, mel, options)

answered Jun 27, 2023 at 12:18

mb7

14 bronze badges

Comments

Brady Dowling · Accepted Answer · 2023-06-28 19:41:19Z

-1

Looks like the documentation/API may have changed since you posted this. The docs now say File uploads are currently limited to 25 MB, which could be an issue for you. The recommendation in the docs is to use openai and open before passing the file to the API, as such:

import openai
audio_file= open("/path/to/file/audio.mp3", "rb")
transcript = openai.Audio.transcribe("whisper-1", audio_file)

They also recommend using AudioSegment to break up the audio file in pieces.

answered Jun 28, 2023 at 19:41

Brady Dowling

5,6325 gold badges41 silver badges67 bronze badges

Collectives™ on Stack Overflow

How to successfully transcribe audio files using Whisper for OpenAI in Python?

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related