gem "elevenlabs"

require "elevenlabs"

# 1. Configure the gem globally (Optional)
Elevenlabs.configure do |config|
  config.api_key = "YOUR_API_KEY"
end

# 2. Initialize a client (will use configured API key)
client = Elevenlabs::Client.new

# 3. List available voices
voices = client.list_voices
puts voices # JSON response with voices

# 4. Convert text to speech
voice_id = "YOUR_VOICE_ID"
text = "Hello from Elevenlabs!"
audio_data = client.text_to_speech(voice_id, text)

# 5. Save the audio file
File.open("output.mp3", "wb") { |f| f.write(audio_data) }
puts "Audio file saved to output.mp3"

# 6. Design a voice with a text prompt
response = client.design_voice(
  "A deep, resonant male voice with a British accent, suitable for storytelling",
  output_format: "mp3_44100_192",
  model_id: "eleven_multilingual_ttv_v2",
  text: "In a land far away, where the mountains meet the sky, a great adventure began. Brave heroes embarked on a quest to find the lost artifact, facing challenges and forging bonds that would last a lifetime. Their journey took them through enchanted forests, across raging rivers, and into the heart of ancient ruins.",
  auto_generate_text: false,
  loudness: 0.5,
  seed: 12345,
  guidance_scale: 5.0,
  stream_previews: false
)

# 7. Save voice preview audio
require "base64"
response["previews"].each_with_index do |preview, index|
  audio_data = Base64.decode64(preview["audio_base_64"])
  File.open("preview_#{index}.mp3", "wb") { |f| f.write(audio_data) }
  puts "Saved preview #{index + 1} to preview_#{index}.mp3"
end

client = Elevenlabs::Client.new(api_key: "DIFFERENT_API_KEY")

# config/initializers/elevenlabs.rb
require "elevenlabs"

Rails.application.config.to_prepare do
  Elevenlabs.configure do |config|
    config.api_key = Rails.application.credentials.dig(:eleven_labs, :api_key)
  end
end

client = Elevenlabs::Client.new

class AudioController < ApplicationController
  def generate
    client = Elevenlabs::Client.new
    voice_id = params[:voice_id]
    text = params[:text]

    begin
      audio_data = client.text_to_speech(voice_id, text)
      send_data audio_data, type: "audio/mpeg", disposition: "attachment", filename: "output.mp3"
    rescue Elevenlabs::APIError => e
      render json: { error: e.message }, status: :bad_request
    end
  end
end

client.list_voices
# => { "voices" => [...] }

2. List Models

client.list_models
# => [...]

3. **Get Voice Details**

```ruby
client.get_voice("VOICE_ID")
# => { "voice_id" => "...", "name" => "...", ... }

sample_files = [File.open("sample1.mp3", "rb")]
client.create_voice("Custom Voice", sample_files, description: "My custom AI voice")
# => JSON response with new voice details

sample_files = [File.open("trump.mp3", "rb")]
client.create_voice("Donald Trump", sample_files, description: "My Trump Voice")
# => {"voice_id"=>"<RETURNED_VOICE_ID>", "requires_verification"=>false}
trump = "<RETURNED_VOICE_ID>"
client.banned?(trump)
# => true

client.edit_voice("VOICE_ID", name: "Updated Voice Name")
# => JSON response with updated details

client.delete_voice("VOICE_ID")
# => JSON response acknowledging deletion

audio_data = client.text_to_speech("VOICE_ID", "Hello world!")
File.open("output.mp3", "wb") { |f| f.write(audio_data) }

IO.popen("play -t mp3 -", "wb") do |audio_pipe| # Notice "wb" (write binary)
  client.text_to_speech_stream("VOICE_ID", "Some text to stream back in chunks") do |chunk|
    audio_pipe.write(chunk.b) # Ensure chunk is written as binary
  end
end

design_response = client.design_voice(
  "A warm, friendly female voice with a slight Australian accent",
  model_id: "eleven_multilingual_ttv_v2",
  text: "Welcome to our podcast, where every story is an adventure, taking you on a journey through fascinating worlds, inspiring voices, and unforgettable moments.",
  auto_generate_text: false
)

generated_voice_id = design_response["previews"].first["generated_voice_id"] #three previews are given, but for this example we will use the first to create a voice here

# Step 2: Create the permanent voice
create_response = client.create_from_generated_voice(
  "Friendly Aussie",
  "A warm, friendly Australian-accented voice for podcasts",
   generated_voice_id,
)

voice_id = create_response["voice_id"] # This is the ID you can use for TTS

# Step 3: Use the new voice for TTS
audio_data = client.text_to_speech(voice_id, "This is my new permanent designed voice.")
File.open("friendly_aussie.mp3", "wb") { |f| f.write(audio_data) }

inputs = [{text: "It smells like updog in here", voice_id: "TX3LPaxmHKxFdv7VOQHJ"}, {text: "What's updog?", voice_id: "RILOU7YmBhvwJGDGjNmP"}, {text: "Not much, you?", voice_id: "TX3LPaxmHKxFdv7VOQHJ"}]

audio_data = client.text_to_dialogue(inputs)
File.open("what's updog.mp3", "wb") { |f| f.write(audio_data) }

audio = client.compose_music(prompt: "Lo-fi hip hop beat", music_length_ms: 30000)
File.binwrite("lofi.mp3", audio)

File.open("epic_stream.mp3", "wb") do |f|
  client.compose_music_stream(prompt: "Epic orchestral build", music_length_ms: 60000) do |chunk|
    f.write(chunk)
  end
end

result = client.compose_music_detailed(prompt: "Jazz piano trio", music_length_ms: 20000)
puts result # raw multipart data (needs parsing)

plan = client.create_music_plan(prompt: "Upbeat pop song with verse and chorus", music_length_ms: 60000)
puts plan[:sections]

audio_data = client.sound_generation("Futuristic laser blast in a space battle")

# Save the audio to a file
File.open("laser_blast.mp3", "wb") { |f| f.write(audio_data) }

audio_data = client.sound_generation(
  "Roaring dragon in a fantasy cave",
  duration_seconds: 3.0,
  prompt_influence: 0.7, # Higher influence for closer adherence to the prompt
  output_format: "mp3_22050_32"
)

# Save the audio to a file
File.open("dragon_roar.mp3", "wb") { |f| f.write(audio_data) }

audio_data = client.sound_generation(
  "Eerie wind and distant owl hoots in a haunted forest",
  loop: true,
  duration_seconds: 10.0,
  prompt_influence: 0.5,
  output_format: "mp3_22050_32"
)
# Save the audio to a file
File.open("haunted_forest_loop.mp3", "wb") { |f| f.write(audio_data) }

begin
  client.design_voice("Short description") # Too short, will raise error
rescue Elevenlabs::UnprocessableEntityError => e
  puts "Validation error: #{e.message}"
rescue Elevenlabs::AuthenticationError => e
  puts "Invalid API key: #{e.message}"
rescue Elevenlabs::NotFoundError => e
  puts "Voice not found: #{e.message}"
rescue Elevenlabs::APIError => e
  puts "General error: #{e.message}"
end