Using Whisper API to Generate .SRT Transcripts? – Whisper

The Problem:

Given an audio file, how can I use the Whisper API to generate an SRT file of transcriptions, allowing speech-to-text functionality without having to run the model locally?

The Solutions:

Solution 1: Generating SRT Transcripts Using OpenAI’s Whisper API

Generating SRT transcripts from audio files is feasible using OpenAI’s Whisper API. By setting the `response_format` parameter to “srt” while creating a transcription, you can obtain the transcript directly in the SRT (SubRip Text) format, which is commonly used for subtitles. Here’s an improved solution:

import os
import tempfile
import openai

openai.api_key = "YOUR_API_KEY"

# Read the audio file
audio_file_path = "path/to/audio_file.wav"

# Optionally, you can set the audio format
audio_format = "wav"  # Supported formats: "mp3", "wav"

# For longer audio files, you may need to store the transcript temporarily
temp_file = tempfile.NamedTemporaryFile(delete=False)

# Perform transcription and store the result in the temporary file
transcription = openai.Audio.transcribe(
    model="whisper",
    file=audio_file_path,
    audio_format=audio_format,
    response_format="srt",
    output_file=temp_file.name,
)

# Get the transcript from the temporary file
transcript_text = open(temp_file.name, "r").read()

# Clean up the temporary file
os.unlink(temp_file.name)

# Do something with the transcript text, e.g., save it or display it
print(transcript_text)

This revised solution includes additional comments and explanations, clarifying the purpose of each step. It also handles storing the transcript in a temporary file, which might be necessary for longer audio files to prevent memory issues. The preferred audio format can also be specified explicitly if desired. Finally, the transcript is printed as an example of how to use the obtained result.

Video Explanation:

The following video, titled "I Made an App that Accurately Produces Subtitles Using Whisper ...", provides additional insights and in-depth exploration related to the topics discussed in this post.

Get the Transcriber: https://github.com/JarodMica/youtube-transcriber Python - https://www.python.org/downloads/ Whisper ...