Whisper+PortAudio

run.jl

using PortAudio, SampledSignals, Whisper

# Requires https://github.com/aviks/Whisper.jl/pull/8
ctx = Whisper.create_ctx("base.en")
wparams = Whisper.create_params()
wparams.n_threads = 6

queue = Channel(typemax(Int))
stream = PortAudioStream(1, 0)
function to_16k(buf)
    buf_16k = SampleBuf(Float32, 16000, round(Int, length(buf)*(16000/samplerate(buf))), nchannels(buf))
    write(SampleBufSink(buf_16k), SampleBufSource(buf))
    return buf_16k
end
errormonitor(Threads.@spawn begin
    fullbuf_16k = to_16k(read(stream, 5s)).data[:]
    while true
        buf = read(stream, 4s)
        buf_16k = to_16k(buf)
        append!(fullbuf_16k, buf_16k.data[:])
        deleteat!(fullbuf_16k, 1:length(buf_16k))
        put!(queue, copy(fullbuf_16k))
    end
end)
full_results = ""
function merge_results!(result)
    result_lower = lowercase(result)
    global full_results
    if endswith(lowercase(full_results), lowercase(result))
        return "BAD THINGS HAPPENED "
    end
    full_results *= " " * result
    return result
end
while true
    samp = take!(queue)
    result = redirect_stderr(devnull) do
        Whisper.transcribe(ctx, wparams, samp)
    end
    diff_results = merge_results!(result)
    print(diff_results)
end