feat: remove espeak, add random voice/style selection, fix punctuation in TTS

- Removed espeak audio provider completely, now only uses OpenAI TTS
- Audio now uses random voice selection by default (can override with --openai-voice)
- Added --all-voices flag to generate audio in all 11 OpenAI voices
- Images now use random art styles (13 different styles including superhero, yoga, cat-themed)
- Fixed TTS to remove punctuation marks before speaking
- Updated Bulgarian pronunciation instructions to explicitly avoid Russian accent

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

15 files changed, 158 insertions, 758 deletions

diff --git a/.totalrecall.yaml.example b/.totalrecall.yaml.example
index e41b97f..e649bde 100644
--- a/.totalrecall.yaml.example
+++ b/.totalrecall.yaml.example
@@ -1,21 +1,14 @@
-# TotalRecall Configuration Example
+# TotalRecalooConfiguration Example
 # Copy this to ~/.totalrecall.yaml or ./.totalrecall.yaml
 
 # Audio configuration
 audio:
-  # Provider: espeak or openai
-  provider: openai
+  # Audio is generated using OpenAI TTS
+  # Provider field removed - OpenAI is now the only option
   
   # Audio output format
   format: mp3
   
-  # ESpeak settings
-  voice: bg+f1
-  speed: 150
-  pitch: 50
-  amplitude: 100
-  word_gap: 0
-  
   # OpenAI TTS settings
   openai_key: ${OPENAI_API_KEY}  # Can also use environment variable
   openai_model: gpt-4o-mini-tts  # Options: tts-1, tts-1-hd, gpt-4o-mini-tts
@@ -24,12 +17,12 @@ audio:
   
   # Voice instructions for gpt-4o-mini-tts model
   # This allows you to customize how the AI speaks
-  openai_instruction: "Speak slowly and clearly with natural Bulgarian pronunciation, emphasizing each syllable distinctly"
+  openai_instruction: "You are speaking Bulgarian language (български език). Pronounce the Bulgarian text with authentic Bulgarian phonetics, not Russian. Speak slowly and clearly for language learners."
   
   # Alternative instruction examples:
-  # openai_instruction: "Pronounce with a native Bulgarian accent, speaking at a moderate pace suitable for language learning"
-  # openai_instruction: "Speak as a patient Bulgarian language teacher, clearly articulating each sound"
-  # openai_instruction: "Use clear Bulgarian pronunciation with slight pauses between syllables"
+  # openai_instruction: "Speak in Bulgarian (not Russian!). Use native Bulgarian pronunciation with clear articulation for each syllable."
+  # openai_instruction: "You are a Bulgarian language teacher. Pronounce the Bulgarian words slowly with authentic Bulgarian accent and phonetics."
+  # openai_instruction: "Speak Bulgarian text with proper Bulgarian pronunciation. Avoid Russian accent. Speak clearly at a pace suitable for beginners."
   
   # Caching
   enable_cache: true
@@ -56,4 +49,4 @@ image:
 
 # Output configuration
 output:
-  directory: ./anki_cards
\ No newline at end of file
+  directory: ./anki_cards
diff --git a/CLAUDE.md b/CLAUDE.md
index 4b972ef..4ab7679 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -6,7 +6,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 **totalrecall** - Bulgarian Anki Flashcard Generator
 
 A Go CLI tool that generates Anki flashcard materials from Bulgarian words:
-- Generates audio pronunciation using espeak-ng
+- Generates audio pronunciation using OpenAI TTS
 - Downloads representative images via web search
 - Creates Anki-compatible output files
 
@@ -60,7 +60,7 @@ golangci-lint run
 totalrecall/
 ├── cmd/totalrecall/          # CLI entry point
 ├── internal/          # Private packages
-│   ├── audio/        # Audio generation (espeak-ng wrapper)
+│   ├── audio/        # Audio generation (OpenAI TTS)
 │   ├── image/        # Image search functionality
 │   ├── anki/         # Anki format generation
 │   ├── config/       # Configuration management
@@ -68,20 +68,13 @@ totalrecall/
 ```
 
 ### Key Design Decisions
-1. **espeak-ng for TTS**: Open source, supports Bulgarian, no API keys needed
-2. **Modular image search**: Support multiple providers (Pixabay, Unsplash)
+1. **OpenAI TTS**: High-quality, natural-sounding Bulgarian pronunciation
+2. **Modular image search**: Support multiple providers (Pixabay, Unsplash, OpenAI DALL-E)
 3. **Configuration via YAML**: User-friendly configuration with viper
 4. **Cobra for CLI**: Industry-standard CLI framework
 
 ### External Dependencies
-- **espeak-ng**: Must be installed on the system
-  ```bash
-  # Ubuntu/Debian
-  sudo apt-get install espeak-ng
-  
-  # macOS
-  brew install espeak-ng
-  ```
+- **OpenAI API Key**: Required for audio generation
 
 ### API Configuration
 Image search APIs require configuration in `.bulg.yaml`:
@@ -89,24 +82,12 @@ Image search APIs require configuration in `.bulg.yaml`:
 - **Unsplash**: Required API key
 
 ## Testing Approach
-1. Unit tests mock external commands (espeak-ng) and API calls
+1. Unit tests mock API calls
 2. Integration tests use real services when available
 3. Test with common Bulgarian words: ябълка, котка, куче, хляб
 
 ## Common Issues and Solutions
 
-### espeak-ng Bulgarian pronunciation
-There have been reported issues with Bulgarian pronunciation in espeak-ng v1.49.3. If pronunciation sounds wrong, try:
-```bash
-# Check version
-espeak-ng --version
-
-# Test Bulgarian voice
-espeak-ng -v bg "Здравей"
-
-# Try different voice variants
-espeak-ng -v bg+f1 "Здравей"
-```
 
 ### Package Declaration Error
 If you see an error about `package main`, ensure `cmd/totalrecall/main.go` has:
@@ -125,4 +106,4 @@ package main  // NOT package bulg
 ## Bulgarian Language Notes
 - Input should be in Cyrillic script
 - Common test words: ябълка (apple), котка (cat), куче (dog)
-- Voice variants: bg+m1 (male), bg+f1 (female)
\ No newline at end of file
+- OpenAI voices: nova, alloy, echo, shimmer (work well for Bulgarian)
\ No newline at end of file
diff --git a/GPT4O_AUDIO_NOTE.md b/GPT4O_AUDIO_NOTE.md
deleted file mode 100644
index a227c60..0000000
--- a/GPT4O_AUDIO_NOTE.md
+++ /dev/null
@@ -1,35 +0,0 @@
-# GPT-4o Audio Support Note
-
-## Current Status
-
-The standard OpenAI Text-to-Speech API (`/v1/audio/speech`) currently supports only:
-- `tts-1` - Standard quality
-- `tts-1-hd` - High definition quality
-
-## GPT-4o Audio Capabilities
-
-According to OpenAI documentation, GPT-4o models have audio capabilities, but these work differently:
-
-1. **Realtime API**: GPT-4o audio generation might be part of the new Realtime API, which uses WebSockets for bidirectional audio streaming.
-
-2. **Chat Completions with Audio**: GPT-4o might support audio output through the chat completions API with special modality parameters, but this requires different request/response handling than the standard TTS API.
-
-3. **Model Names**: Models like `gpt-4o-audio-preview` or `gpt-4o-mini` with audio capabilities might not be compatible with the standard TTS endpoint.
-
-## Experimental Usage
-
-You can try experimental model names with the `--openai-model` flag:
-```bash
-./totalrecall "word" --openai-model gpt-4o-audio-preview
-```
-
-However, this will likely result in a 404 error as these models require different API endpoints.
-
-## Future Implementation
-
-To properly support GPT-4o audio generation, we would need to:
-1. Implement support for the Realtime API (WebSocket-based)
-2. Or implement the chat completions API with audio modalities
-3. Handle different request/response formats for audio data
-
-For now, stick with `tts-1` or `tts-1-hd` for reliable audio generation.
\ No newline at end of file
diff --git a/README.md b/README.md
index c866e23..7177b66 100644
--- a/README.md
+++ b/README.md
@@ -4,46 +4,28 @@
 
 It has mainly been vibe coded using Claude Code CLI.
 
-⚠️ **Important:** This tool uses OpenAI services by default, which requires an API key. See [Quick Start](#quick-start) for setup instructions or use the free alternatives with `--audio-provider espeak --image-api pixabay`.
+⚠️ **Important:** This tool uses OpenAI services for audio generation, which requires an API key. See [Quick Start](#quick-start) for setup instructions.
 
 ## Features
 
-- Audio generation with multiple providers:
-  - **espeak-ng**: Free, offline Bulgarian voices (robotic quality)
-  - **OpenAI TTS**: High-quality, natural-sounding voices (requires API key)
+- Audio generation using **OpenAI TTS**: High-quality, natural-sounding voices (requires API key)
+  - Random voice selection by default for variety
+  - Option to generate in all 11 available voices
 - Image search and generation:
   - **Pixabay**: Free stock photo search (optional API key)
   - **Unsplash**: High-quality photo search (requires API key)
-  - **OpenAI DALL-E**: AI-generated educational images (requires API key)
+  - **OpenAI DALL-E**: AI-generated educational images with random art styles (requires API key)
 - Batch processing of multiple words
 - Anki-compatible CSV export
 - Configurable voice variants and speech speed
 - Support for WAV and MP3 audio formats
-- Audio caching to save API costs (OpenAI)
+- Audio and image caching to save API costs
 
 ## Installation
 
 ### Prerequisites
 
-1. **For espeak-ng audio** (free, offline):
-   ```bash
-   # Ubuntu/Debian
-   sudo apt-get install espeak-ng
-   
-   # macOS
-   brew install espeak-ng
-   ```
-
-2. **ffmpeg** (optional, for MP3 conversion with espeak):
-   ```bash
-   # Ubuntu/Debian
-   sudo apt-get install ffmpeg
-   
-   # macOS
-   brew install ffmpeg
-   ```
-
-3. **For OpenAI TTS** (paid, high quality):
+1. **For OpenAI TTS** (required for audio generation):
    - Create an account at https://platform.openai.com
    - Generate an API key at https://platform.openai.com/api-keys
    - Set the key using one of these methods:
@@ -76,9 +58,9 @@ export OPENAI_API_KEY="sk-..."
    totalrecall ябълка
    ```
 
-2. Use free alternatives (espeak + pixabay):
+2. Use free Pixabay for images:
    ```bash
-   totalrecall ябълка --audio-provider espeak --image-api pixabay
+   totalrecall ябълка --image-api pixabay
    ```
 
 3. Process multiple words from a file:
@@ -97,20 +79,14 @@ Create a `.totalrecall.yaml` file in your home directory or project folder:
 
 ```yaml
 audio:
-  provider: openai       # Audio provider (espeak or openai) - default: openai
   format: mp3           # Audio format (wav or mp3)
   
-  # ESpeak settings
-  voice: bg+f1          # Voice variant (bg, bg+m1, bg+f1, etc.)
-  speed: 150            # Speech speed (80-450 words/minute)
-  pitch: 50             # Pitch adjustment (0-99)
-  
   # OpenAI settings
   openai_key: "sk-..."  # Your OpenAI API key
   openai_model: "gpt-4o-mini-tts" # Model: tts-1, tts-1-hd, or gpt-4o-mini-tts
   openai_voice: "nova"  # Voice: alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse
   openai_speed: 0.8     # Speed: 0.25 to 4.0 (may be ignored by gpt-4o-mini models)
-  openai_instruction: "Speak slowly and clearly with natural Bulgarian pronunciation" # For gpt-4o-mini models only
+  openai_instruction: "You are speaking Bulgarian language (български език). Pronounce the Bulgarian text with authentic Bulgarian phonetics, not Russian." # For gpt-4o-mini models only
   
   # Caching
   enable_cache: true
@@ -153,18 +129,13 @@ totalrecall [word] [flags]
 - `--skip-images`: Skip image download
 - `--images-per-word int`: Number of images per word (default 1)
 - `--image-api string`: Image source - pixabay, unsplash, or openai (default "openai")
+- `--all-voices`: Generate audio in all available OpenAI voices (creates 11 files per word)
 
-#### Audio Provider Options
-- `--audio-provider string`: Audio provider - espeak or openai (default "openai")
-
-#### ESpeak Tuning Options
-- `--pitch int`: Pitch adjustment 0-99 (default 50, lower=deeper, espeak only)
-- `--amplitude int`: Volume 0-200 (default 100, espeak only)
-- `--word-gap int`: Gap between words in 10ms units (default 0, espeak only)
+#### Audio Options
 
 #### OpenAI Audio Options
 - `--openai-model string`: Model - tts-1, tts-1-hd, or gpt-4o-mini-tts (default "gpt-4o-mini-tts", requires special access)
-- `--openai-voice string`: Voice - alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse (default "nova")
+- `--openai-voice string`: Voice - alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse (default: random)
 - `--openai-speed float`: Speech speed 0.25-4.0 (default 0.8, may be ignored by gpt-4o-mini-tts)
 - `--openai-instruction string`: Voice instructions for gpt-4o-mini-tts model (e.g., "speak with a Bulgarian accent")
 
@@ -197,11 +168,8 @@ totalrecall [word] [flags]
 # Single word (uses OpenAI by default)
 totalrecall котка
 
-# Using espeak-ng (free alternative)
-totalrecall котка --audio-provider espeak
-
 # High-quality OpenAI with specific voice
-totalrecall ябълка --audio-provider openai --openai-model tts-1-hd --openai-voice alloy
+totalrecall ябълка --openai-model tts-1-hd --openai-voice alloy
 
 # Use gpt-4o-mini-tts with custom voice instructions
 totalrecall ябълка --openai-instruction "Speak like a patient Bulgarian teacher, very slowly and clearly"
@@ -209,9 +177,6 @@ totalrecall ябълка --openai-instruction "Speak like a patient Bulgarian te
 # Multiple words with custom output
 totalrecall --batch animals.txt -o ./animal_cards
 
-# ESpeak with tuning
-totalrecall ябълка --pitch 40 --word-gap 3
-
 # Skip images, audio only
 totalrecall куче --skip-images
 
@@ -225,7 +190,10 @@ totalrecall ябълка --image-api openai
 totalrecall котка --image-api openai --openai-image-model dall-e-3 --openai-image-quality hd
 
 # Combine OpenAI audio and images
-totalrecall куче --audio-provider openai --image-api openai
+totalrecall куче --image-api openai
+
+# Generate audio in all 11 OpenAI voices
+totalrecall котка --all-voices --skip-images
 ```
 
 ### Batch File Format
@@ -255,8 +223,6 @@ Available Bulgarian voices:
 
 ## Troubleshooting
 
-### espeak-ng not found
-Make sure espeak-ng is installed and in your PATH.
 
 ### No images found
 - Check your internet connection
@@ -275,48 +241,28 @@ Make sure espeak-ng is installed and in your PATH.
 - Both services cache results to avoid regenerating identical content
 
 ### Free Alternatives
-- **Audio**: Use espeak-ng (free but robotic quality)
 - **Images**: Use Pixabay without API key (limited rate)
 
 ### OpenAI Troubleshooting
 - Check the API key has proper permissions enabled
 - If you get rate limit errors, wait a moment and try again
-- The tool will automatically fall back to espeak-ng if OpenAI audio fails
-
-### Audio sounds robotic
-The Bulgarian voice in espeak-ng can sound robotic. To improve quality:
-
-```bash
-# Test with different settings
-espeak-ng -v bg -p 40 -s 140 "Здравей"  # Deeper, slower
-espeak-ng -v bg+f1 -p 60 -g 2 "Здравей"  # Higher pitch, word gaps
-
-# Using totalrecall with tuning
-totalrecall ябълка --pitch 40 --word-gap 2 --amplitude 120
-```
 
-Recommended settings for clearer pronunciation:
-- `--pitch 40`: Slightly deeper voice (less robotic)
-- `--word-gap 2-5`: Small gaps between words
-- `--amplitude 120`: Slightly louder
-- `-v bg+f1`: Female variant often sounds clearer
 
-### Using OpenAI for Better Quality
+### OpenAI TTS Configuration
 
-OpenAI TTS provides much more natural Bulgarian pronunciation:
+OpenAI TTS provides natural Bulgarian pronunciation:
 
 ```bash
 # Option 1: Use environment variable
 export OPENAI_API_KEY="sk-your-key-here"
-totalrecall ябълка --audio-provider openai
+totalrecall ябълка
 
 # Option 2: Set in .totalrecall.yaml
 audio:
-  provider: openai
   openai_key: "sk-your-key-here"
 
 # Use with custom voice
-totalrecall ябълка --audio-provider openai --openai-voice alloy
+totalrecall ябълка --openai-voice alloy
 ```
 
 **OpenAI TTS Models**:
diff --git a/TODO.md b/TODO.md
index 4ceec03..9b41e0a 100644
--- a/TODO.md
+++ b/TODO.md
@@ -1,18 +1,2 @@
 # TODO's
 
-## Completed
-1.  [x] Implement OpenAI DALL-E image generation for flashcards
-    - [x] Create OpenAI image provider implementing ImageSearcher interface
-    - [x] Add configuration flags for DALL-E model, size, quality, and style
-    - [x] Implement caching mechanism to avoid regenerating identical images
-    - [x] Create educational prompt generation for language learning
-    - [x] Add OpenAI provider to image download workflow
-    - [x] Update documentation with examples and configuration
-
-## In Progress / Remaining
-1.  [ ] Write unit tests for OpenAI image provider
-2.  [ ] Add cost estimation warnings in output (show estimated API costs)
-3.  [ ] Test with common Bulgarian words (ябълка, котка, куче, хляб)
-4.  [ ] Consider adding batch image generation for cost optimization
-5.  [ ] Add image style presets for different learning contexts (e.g., children, adults)
-6.  [ ] Implement fallback from OpenAI to other providers on failure
diff --git a/cmd/totalrecall/main.go b/cmd/totalrecall/main.go
index df9e666..2a43d51 100644
--- a/cmd/totalrecall/main.go
+++ b/cmd/totalrecall/main.go
@@ -3,10 +3,12 @@ package main
 import (
 	"context"
 	"fmt"
+	"math/rand"
 	"os"
 	"path/filepath"
 	"sort"
 	"strings"
+	"time"
 
 	"github.com/sashabaranov/go-openai"
 	"github.com/spf13/cobra"
@@ -21,7 +23,7 @@ import (
 var (
 	// Flags
 	cfgFile      string
-	voice        string
+	// voice removed - was only for espeak
 	outputDir    string
 	audioFormat  string
 	imageAPI     string
@@ -31,12 +33,8 @@ var (
 	imagesPerWord int
 	generateAnki bool
 	listModels   bool
-	// Audio provider flags
-	audioProvider  string
-	// Audio tuning flags (espeak)
-	audioPitch     int
-	audioAmplitude int
-	audioWordGap   int
+	allVoices    bool
+	// Audio provider flags removed - now only OpenAI
 	// OpenAI flags
 	openAIModel       string
 	openAIVoice       string
@@ -55,7 +53,7 @@ var rootCmd = &cobra.Command{
 	Short: "Bulgarian Anki Flashcard Generator",
 	Long: `totalrecall generates Anki flashcard materials from Bulgarian words.
 
-It creates audio pronunciation files using espeak-ng and downloads
+It creates audio pronunciation files using OpenAI TTS and downloads
 representative images from web search APIs.
 
 Example:
@@ -69,11 +67,13 @@ Example:
 func init() {
 	cobra.OnInitialize(initConfig)
 	
+	// Initialize random number generator
+	rand.Seed(time.Now().UnixNano())
+	
 	// Global flags
 	rootCmd.PersistentFlags().StringVar(&cfgFile, "config", "", "config file (default is $HOME/.totalrecall.yaml)")
 	
 	// Local flags
-	rootCmd.Flags().StringVarP(&voice, "voice", "v", "bg+f1", "Voice variant (bg, bg+m1, bg+f1, etc.)")
 	rootCmd.Flags().StringVarP(&outputDir, "output", "o", "./anki_cards", "Output directory")
 	rootCmd.Flags().StringVarP(&audioFormat, "format", "f", "mp3", "Audio format (wav or mp3)")
 	rootCmd.Flags().StringVar(&imageAPI, "image-api", "openai", "Image source (pixabay, unsplash, or openai)")
@@ -83,18 +83,13 @@ func init() {
 	rootCmd.Flags().IntVar(&imagesPerWord, "images-per-word", 1, "Number of images to download per word")
 	rootCmd.Flags().BoolVar(&generateAnki, "anki", false, "Generate Anki import CSV file")
 	rootCmd.Flags().BoolVar(&listModels, "list-models", false, "List available OpenAI models for the current API key")
+	rootCmd.Flags().BoolVar(&allVoices, "all-voices", false, "Generate audio in all available voices (creates multiple files)")
 	
-	// Audio provider selection
-	rootCmd.Flags().StringVar(&audioProvider, "audio-provider", "openai", "Audio provider: espeak or openai")
-	
-	// Audio tuning flags (espeak)
-	rootCmd.Flags().IntVar(&audioPitch, "pitch", 50, "Audio pitch adjustment (0-99, default 50, espeak only)")
-	rootCmd.Flags().IntVar(&audioAmplitude, "amplitude", 100, "Audio volume (0-200, default 100, espeak only)")
-	rootCmd.Flags().IntVar(&audioWordGap, "word-gap", 0, "Gap between words in 10ms units (default 0, espeak only)")
+	// Audio provider removed - now only OpenAI
 	
 	// OpenAI flags
 	rootCmd.Flags().StringVar(&openAIModel, "openai-model", "gpt-4o-mini-tts", "OpenAI TTS model: tts-1, tts-1-hd, gpt-4o-mini-tts")
-	rootCmd.Flags().StringVar(&openAIVoice, "openai-voice", "nova", "OpenAI voice: alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse")
+	rootCmd.Flags().StringVar(&openAIVoice, "openai-voice", "", "OpenAI voice: alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse (default: random)")
 	rootCmd.Flags().Float64Var(&openAISpeed, "openai-speed", 0.8, "OpenAI speech speed (0.25 to 4.0, may be ignored by gpt-4o-mini-tts)")
 	rootCmd.Flags().StringVar(&openAIInstruction, "openai-instruction", "", "Voice instructions for gpt-4o-mini-tts model (e.g., 'speak slowly with a Bulgarian accent')")
 	
@@ -237,23 +232,47 @@ func processWord(word string) error {
 }
 
 func generateAudio(word string) error {
+	allVoicesList := []string{"alloy", "ash", "ballad", "coral", "echo", "fable", "onyx", "nova", "sage", "shimmer", "verse"}
+	
+	// Get list of voices to use
+	var voices []string
+	if allVoices {
+		voices = allVoicesList
+	} else if openAIVoice != "" {
+		// Use explicitly specified voice
+		voices = []string{openAIVoice}
+		fmt.Printf("  Using specified voice: %s\n", openAIVoice)
+	} else {
+		// Select a random voice
+		randomVoice := allVoicesList[rand.Intn(len(allVoicesList))]
+		voices = []string{randomVoice}
+		fmt.Printf("  Using random voice: %s\n", randomVoice)
+	}
+	
+	// Generate audio for each voice
+	for i, voice := range voices {
+		if allVoices {
+			fmt.Printf("  Generating audio %d/%d (voice: %s)...\n", i+1, len(voices), voice)
+		}
+		if err := generateAudioWithVoice(word, voice); err != nil {
+			return fmt.Errorf("failed to generate audio with voice %s: %w", voice, err)
+		}
+	}
+	
+	return nil
+}
+
+func generateAudioWithVoice(word, voice string) error {
 	// Create audio provider configuration
 	providerConfig := &audio.Config{
-		Provider:     audioProvider,
+		Provider:     "openai",
 		OutputDir:    outputDir,
 		OutputFormat: audioFormat,
 		
-		// ESpeak settings
-		ESpeakVoice:     voice,
-		ESpeakSpeed:     viper.GetInt("audio.speed"),
-		ESpeakPitch:     audioPitch,
-		ESpeakAmplitude: audioAmplitude,
-		ESpeakWordGap:   audioWordGap,
-		
 		// OpenAI settings
 		OpenAIKey:         getOpenAIKey(),
 		OpenAIModel:       openAIModel,
-		OpenAIVoice:       openAIVoice,
+		OpenAIVoice:       voice,
 		OpenAISpeed:       openAISpeed,
 		OpenAIInstruction: openAIInstruction,
 		
@@ -263,26 +282,11 @@ func generateAudio(word string) error {
 	}
 	
 	// Set defaults
-	if providerConfig.ESpeakSpeed == 0 {
-		providerConfig.ESpeakSpeed = 150
-	}
 	if providerConfig.CacheDir == "" {
 		providerConfig.CacheDir = "./.audio_cache"
 	}
 	
 	// Use config file values if not overridden by flags
-	if audioProvider == "openai" && viper.IsSet("audio.provider") {
-		providerConfig.Provider = viper.GetString("audio.provider")
-	}
-	if audioPitch == 50 && viper.IsSet("audio.pitch") {
-		providerConfig.ESpeakPitch = viper.GetInt("audio.pitch")
-	}
-	if audioAmplitude == 100 && viper.IsSet("audio.amplitude") {
-		providerConfig.ESpeakAmplitude = viper.GetInt("audio.amplitude")
-	}
-	if audioWordGap == 0 && viper.IsSet("audio.word_gap") {
-		providerConfig.ESpeakWordGap = viper.GetInt("audio.word_gap")
-	}
 	if openAIModel == "gpt-4o-mini-tts" && viper.IsSet("audio.openai_model") {
 		providerConfig.OpenAIModel = viper.GetString("audio.openai_model")
 	}
@@ -299,25 +303,20 @@ func generateAudio(word string) error {
 	// Create the audio provider
 	provider, err := audio.NewProvider(providerConfig)
 	if err != nil {
-		// If OpenAI fails, try to create a fallback to espeak
-		if providerConfig.Provider == "openai" {
-			fmt.Printf("Warning: OpenAI audio provider failed (%v), falling back to espeak-ng\n", err)
-			providerConfig.Provider = "espeak"
-			fallbackProvider, fallbackErr := audio.NewProvider(providerConfig)
-			if fallbackErr != nil {
-				return fmt.Errorf("both OpenAI and espeak-ng failed: %v", fallbackErr)
-			}
-			provider = fallbackProvider
-		} else {
-			return err
-		}
+		return err
 	}
 	
 	// Generate audio file
+	ctx := context.Background()
 	filename := sanitizeFilename(word)
-	outputFile := filepath.Join(outputDir, fmt.Sprintf("%s.%s", filename, audioFormat))
 	
-	ctx := context.Background()
+	// Add voice name to filename if generating multiple voices
+	if allVoices {
+		outputFile := filepath.Join(outputDir, fmt.Sprintf("%s_%s.%s", filename, voice, audioFormat))
+		return provider.GenerateAudio(ctx, word, outputFile)
+	}
+	
+	outputFile := filepath.Join(outputDir, fmt.Sprintf("%s.%s", filename, audioFormat))
 	return provider.GenerateAudio(ctx, word, outputFile)
 }
 
diff --git a/internal/audio/doc.go b/internal/audio/doc.go
index 1fd216b..c8a5ce4 100644
--- a/internal/audio/doc.go
+++ b/internal/audio/doc.go
@@ -1,3 +1,3 @@
-// Package audio provides audio generation functionality using espeak-ng
+// Package audio provides audio generation functionality using OpenAI TTS
 // for Bulgarian text-to-speech conversion.
 package audio
diff --git a/internal/audio/espeak.go b/internal/audio/espeak.go
deleted file mode 100644
index cd42360..0000000
--- a/internal/audio/espeak.go
+++ /dev/null
@@ -1,217 +0,0 @@
-package audio
-
-import (
-	"fmt"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"strings"
-)
-
-// ESpeakConfig holds configuration for espeak-ng audio generation
-type ESpeakConfig struct {
-	Voice     string // Voice variant (e.g., "bg", "bg+m1", "bg+f1")
-	Speed     int    // Speech speed in words per minute (default: 150)
-	Pitch     int    // Pitch adjustment, 0 to 99 (default: 50)
-	Amplitude int    // Volume/amplitude, 0 to 200 (default: 100)
-	WordGap   int    // Gap between words in 10ms units (default: 0)
-	OutputDir string // Directory for output files
-}
-
-// DefaultConfig returns the default configuration for Bulgarian voice
-func DefaultConfig() *ESpeakConfig {
-	return &ESpeakConfig{
-		Voice:     "bg",
-		Speed:     150,
-		Pitch:     50,
-		Amplitude: 100,
-		WordGap:   0,
-		OutputDir: "./",
-	}
-}
-
-// ESpeak provides an interface to the espeak-ng text-to-speech engine
-type ESpeak struct {
-	config *ESpeakConfig
-}
-
-// New creates a new ESpeak instance with the given configuration
-func New(config *ESpeakConfig) (*ESpeak, error) {
-	// Check if espeak-ng is installed
-	if err := checkESpeakInstalled(); err != nil {
-		return nil, err
-	}
-	
-	if config == nil {
-		config = DefaultConfig()
-	}
-	
-	return &ESpeak{config: config}, nil
-}
-
-// GenerateAudio generates an audio file for the given Bulgarian text
-func (e *ESpeak) GenerateAudio(text string, outputFile string) error {
-	// Validate input
-	if text == "" {
-		return fmt.Errorf("text cannot be empty")
-	}
-	
-	// Ensure output directory exists
-	dir := filepath.Dir(outputFile)
-	if dir != "" && dir != "." {
-		if err := os.MkdirAll(dir, 0755); err != nil {
-			return fmt.Errorf("failed to create output directory: %w", err)
-		}
-	}
-	
-	// Build espeak-ng command
-	args := []string{
-		"-v", e.config.Voice,  // Voice selection
-		"-s", fmt.Sprintf("%d", e.config.Speed),  // Speed
-		"-p", fmt.Sprintf("%d", e.config.Pitch),  // Pitch
-		"-a", fmt.Sprintf("%d", e.config.Amplitude),  // Amplitude/volume
-	}
-	
-	// Add word gap if specified
-	if e.config.WordGap > 0 {
-		args = append(args, "-g", fmt.Sprintf("%d", e.config.WordGap))
-	}
-	
-	// Add output file and text
-	args = append(args, "-w", outputFile, text)
-	
-	cmd := exec.Command("espeak-ng", args...)
-	
-	// Run the command
-	output, err := cmd.CombinedOutput()
-	if err != nil {
-		return fmt.Errorf("espeak-ng failed: %w\nOutput: %s", err, string(output))
-	}
-	
-	return nil
-}
-
-// SetVoice updates the voice variant
-func (e *ESpeak) SetVoice(voice string) {
-	e.config.Voice = voice
-}
-
-// SetSpeed updates the speech speed
-func (e *ESpeak) SetSpeed(speed int) {
-	if speed < 80 {
-		speed = 80
-	} else if speed > 450 {
-		speed = 450
-	}
-	e.config.Speed = speed
-}
-
-// SetPitch updates the pitch (0-99, 50 is default)
-func (e *ESpeak) SetPitch(pitch int) {
-	if pitch < 0 {
-		pitch = 0
-	} else if pitch > 99 {
-		pitch = 99
-	}
-	e.config.Pitch = pitch
-}
-
-// SetAmplitude updates the volume/amplitude (0-200, 100 is default)
-func (e *ESpeak) SetAmplitude(amplitude int) {
-	if amplitude < 0 {
-		amplitude = 0
-	} else if amplitude > 200 {
-		amplitude = 200
-	}
-	e.config.Amplitude = amplitude
-}
-
-// SetWordGap updates the gap between words in 10ms units
-func (e *ESpeak) SetWordGap(gap int) {
-	if gap < 0 {
-		gap = 0
-	}
-	e.config.WordGap = gap
-}
-
-// checkESpeakInstalled verifies that espeak-ng is available on the system
-func checkESpeakInstalled() error {
-	cmd := exec.Command("espeak-ng", "--version")
-	if err := cmd.Run(); err != nil {
-		return fmt.Errorf("espeak-ng is not installed or not in PATH: %w", err)
-	}
-	return nil
-}
-
-// ValidateBulgarianText performs basic validation of Bulgarian text
-func ValidateBulgarianText(text string) error {
-	if text == "" {
-		return fmt.Errorf("text cannot be empty")
-	}
-	
-	// Check if text contains at least one Cyrillic character
-	hasCyrillic := false
-	for _, r := range text {
-		// Bulgarian Cyrillic range
-		if (r >= 'А' && r <= 'я') || r == 'Ё' || r == 'ё' {
-			hasCyrillic = true
-			break
-		}
-	}
-	
-	if !hasCyrillic {
-		return fmt.Errorf("text must contain Bulgarian Cyrillic characters")
-	}
-	
-	return nil
-}
-
-// ListVoices returns available Bulgarian voice variants
-func ListVoices() []string {
-	return []string{
-		"bg",      // Default Bulgarian voice
-		"bg+m1",   // Bulgarian male voice 1
-		"bg+m2",   // Bulgarian male voice 2
-		"bg+m3",   // Bulgarian male voice 3
-		"bg+f1",   // Bulgarian female voice 1
-		"bg+f2",   // Bulgarian female voice 2
-		"bg+f3",   // Bulgarian female voice 3
-	}
-}
-
-// ConvertWAVToMP3 converts a WAV file to MP3 using ffmpeg
-func ConvertWAVToMP3(wavFile, mp3File string) error {
-	// Check if ffmpeg is installed
-	if err := exec.Command("ffmpeg", "-version").Run(); err != nil {
-		return fmt.Errorf("ffmpeg is not installed or not in PATH: %w", err)
-	}
-	
-	cmd := exec.Command("ffmpeg", "-i", wavFile, "-acodec", "mp3", "-y", mp3File)
-	output, err := cmd.CombinedOutput()
-	if err != nil {
-		return fmt.Errorf("ffmpeg conversion failed: %w\nOutput: %s", err, string(output))
-	}
-	
-	return nil
-}
-
-// GenerateMP3 generates an MP3 file for the given Bulgarian text
-func (e *ESpeak) GenerateMP3(text string, outputFile string) error {
-	// Generate temporary WAV file
-	tempWAV := strings.TrimSuffix(outputFile, filepath.Ext(outputFile)) + "_temp.wav"
-	
-	// Generate WAV
-	if err := e.GenerateAudio(text, tempWAV); err != nil {
-		return err
-	}
-	
-	// Convert to MP3
-	if err := ConvertWAVToMP3(tempWAV, outputFile); err != nil {
-		// Clean up temporary file
-		os.Remove(tempWAV)
-		return err
-	}
-	
-	// Clean up temporary file
-	return os.Remove(tempWAV)
-}
\ No newline at end of file
diff --git a/internal/audio/espeak_provider.go b/internal/audio/espeak_provider.go
deleted file mode 100644
index 177e2a6..0000000
--- a/internal/audio/espeak_provider.go
+++ /dev/null
@@ -1,65 +0,0 @@
-package audio
-
-import (
-	"context"
-	"path/filepath"
-	"strings"
-)
-
-// ESpeakProvider implements Provider interface for espeak-ng
-type ESpeakProvider struct {
-	espeak *ESpeak
-	format string
-}
-
-// NewESpeakProvider creates a new espeak-ng provider
-func NewESpeakProvider(config *ESpeakConfig) (Provider, error) {
-	espeak, err := New(config)
-	if err != nil {
-		return nil, err
-	}
-	
-	return &ESpeakProvider{
-		espeak: espeak,
-		format: "mp3", // default format
-	}, nil
-}
-
-// GenerateAudio generates audio using espeak-ng
-func (p *ESpeakProvider) GenerateAudio(ctx context.Context, text string, outputFile string) error {
-	// Validate Bulgarian text
-	if err := ValidateBulgarianText(text); err != nil {
-		return err
-	}
-	
-	// Determine format from output file extension
-	ext := strings.ToLower(filepath.Ext(outputFile))
-	
-	switch ext {
-	case ".mp3":
-		return p.espeak.GenerateMP3(text, outputFile)
-	case ".wav":
-		return p.espeak.GenerateAudio(text, outputFile)
-	default:
-		// Default to MP3 if extension is unclear
-		if !strings.HasSuffix(outputFile, ".mp3") {
-			outputFile += ".mp3"
-		}
-		return p.espeak.GenerateMP3(text, outputFile)
-	}
-}
-
-// Name returns the provider name
-func (p *ESpeakProvider) Name() string {
-	return "espeak-ng"
-}