gcp_diarize_test.rb

#!/usr/bin/env ruby
# frozen_string_literal: true

# Speech-to-Text v2 diarization test (chirp_3 model, Japanese by default).
#
# KEY FINDING: chirp_3 + diarizationConfig only works in the "us" multi-region.
# It is rejected everywhere else tested (global, us-central1, europe-west4,
# asia-southeast1, europe-west3) with either "model does not exist" or
# "Permission denied ... It is no longer generally available." Only the
# "us" multi-region endpoint accepts it, and only via the async
# batch_recognize method (sync recognize never supports diarization, in
# v1 or v2, in any region).
#
# Usage:
#   script/gcp_diarize_test.rb <path-to-audio-file> <gcs-bucket-name> [language-code]
#
# Example:
#   script/gcp_diarize_test.rb ./recording.mp3 my-project-scratch-bucket ja-JP
#
# Auth: uses local Application Default Credentials.
#   gcloud auth application-default login --project <PROJECT_ID>
#
# The GCS bucket must already exist in the same project you're calling the
# Speech API against (cross-project GCS reads commonly hit permission
# errors from the Speech service agent).

require "bundler/inline"

gemfile do
  source "https://rubygems.org"
  gem "google-cloud-speech-v2"
end

require "google/cloud/speech/v2"
require "open3"
require "json"

audio_file, bucket, language_code = ARGV
if audio_file.nil? || bucket.nil?
  warn "Usage: #{$PROGRAM_NAME} <audio-file> <gcs-bucket> [language-code]"
  exit 1
end
language_code ||= "ja-JP"

project = `gcloud config get-value project 2>/dev/null`.strip
raise "No active gcloud project. Run: gcloud config set project <PROJECT_ID>" if project.empty?

file_name = File.basename(audio_file)
gcs_audio_uri = "gs://#{bucket}/diarize-test/#{file_name}"
gcs_output_uri = "gs://#{bucket}/diarize-test/output/"

puts "Project:       #{project}"
puts "Audio file:    #{audio_file}"
puts "Language:      #{language_code}"
puts "GCS audio URI: #{gcs_audio_uri}"
puts

puts "==> Uploading audio to GCS..."
system("gcloud", "storage", "cp", audio_file, gcs_audio_uri, "--project", project, exception: true)

client = Google::Cloud::Speech::V2::Speech::Client.new do |config|
  config.endpoint = "us-speech.googleapis.com"
end

recognizer = "projects/#{project}/locations/us/recognizers/_"

request = {
  recognizer: recognizer,
  files: [ { uri: gcs_audio_uri } ],
  config: {
    auto_decoding_config: {},
    model: "chirp_3",
    language_codes: [ language_code ],
    features: {
      enable_automatic_punctuation: true,
      enable_word_time_offsets: true,
      diarization_config: {}
    }
  },
  recognition_output_config: {
    gcs_output_config: { uri: gcs_output_uri }
  }
}

puts "==> Submitting batch_recognize (model=chirp_3, location=us)..."
operation = client.batch_recognize(request)

puts "==> Waiting for operation to complete..."
operation.wait_until_done!
raise "Operation failed: #{operation.results}" if operation.error?

result_uri = operation.response.results.values.first.uri
puts "==> Downloading transcript from #{result_uri}"

json_text, status = Open3.capture2("gcloud", "storage", "cat", result_uri, "--project", project)
raise "Failed to download transcript" unless status.success?

data = JSON.parse(json_text)

puts
puts "==================== DIARIZED TRANSCRIPT ===================="
data.fetch("results", []).each do |r|
  alt = r.fetch("alternatives", []).first
  next unless alt
  words = alt.fetch("words", [])
  next if words.empty?

  current_speaker = nil
  line = []
  words.each do |w|
    speaker = w["speakerLabel"]
    if speaker != current_speaker
      puts "Speaker #{current_speaker}: #{line.join}" unless line.empty?
      current_speaker = speaker
      line = []
    end
    line << w["word"]
  end
  puts "Speaker #{current_speaker}: #{line.join}" unless line.empty?
end
puts "================================================================"