Audio Convert For .ogg For Wisper

[jacob7395]

I know this strictly isn't the place for this but I've been smashing my head into a wall on this. I've got a .ogg file stored in an azure blob. I can't find a solution for converting the file into any of the formats supported Wisper.

I'm sure there is a easy solution to this would really appreciate some help.

[Frogley]

Have you tried FFmpeg?

[jacob7395]

I want to deploy it on azure function so FFmpeg might be a pain to deploy. I managed to pull something together using gpt4. I'll put my solution here then close the discussion.

[jacob7395]

Scraped together with gpt4, tested using the output with whisper and it produce some transcripts.

using Concentus.Oggfile;
using Concentus.Structs;
using NAudio.Lame;

namespace  OpusConverter;

public class OpusStreamToMp3BytesConverter
{
    public static async Task<byte[]> ConvertOpusToMp3(Stream opusStream)
    {
        OpusDecoder opusDecoder = new(48000, 2);
        OpusOggReadStream oggReadStream = new(opusDecoder, opusStream);
        using MemoryStream intermediateStream = new();
        await using (LameMP3FileWriter mp3Writer = new(intermediateStream, new(48000, 16, 2), 128))
        {
            while (oggReadStream.DecodeNextPacket() is { } decodedSamples)
            {
                byte[] buffer = new byte[decodedSamples.Length * sizeof(short)];
                Buffer.BlockCopy(decodedSamples, 0, buffer, 0, buffer.Length);
                mp3Writer.Write(buffer, 0, buffer.Length);
            }

            await mp3Writer.FlushAsync();
        }

        intermediateStream.Position = 0;
        return intermediateStream.ToArray();
    }
}