Harden GUI audio loading for Gemini

6 files changed, 249 insertions, 43 deletions

diff --git a/internal/gui/app.go b/internal/gui/app.go
index a96c627..05955cf 100644
--- a/internal/gui/app.go
+++ b/internal/gui/app.go
@@ -116,7 +116,7 @@ type Config struct {
 	GoogleAPIKey  string
 	// GeminiTTSModel selects the Gemini TTS model when Gemini audio is active.
 	GeminiTTSModel string
-	// GeminiVoice selects a specific Gemini voice; empty uses a random shared voice.
+	// GeminiVoice selects a specific Gemini voice; empty uses the model default.
 	GeminiVoice         string
 	TranslationProvider translation.Provider
 	PhoneticProvider    phonetic.Provider
@@ -1925,6 +1925,11 @@ func (a *Application) clearUI() {
 
 	a.imageDisplay.Clear()
 	a.audioPlayer.Clear()
+	a.currentAudioFile = ""
+	a.currentAudioFileBack = ""
+	a.currentImage = ""
+	a.currentTranslation = ""
+	a.currentCardType = ""
 	// Don't clear the word input or translation entry - they should stay populated
 	// Clear the image prompt entry - it will be loaded from disk if available
 	a.imagePromptEntry.SetText("")
diff --git a/internal/gui/audio_paths.go b/internal/gui/audio_paths.go
new file mode 100644
index 0000000..baee232
--- /dev/null
+++ b/internal/gui/audio_paths.go
@@ -0,0 +1,74 @@
+package gui
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+)
+
+var supportedAudioExtensions = map[string]struct{}{
+	".aac":  {},
+	".flac": {},
+	".mp3":  {},
+	".opus": {},
+	".wav":  {},
+}
+
+func (a *Application) resolveSingleAudioFile(wordDir string) string {
+	return resolveAudioFileByBaseName(wordDir, "audio")
+}
+
+func (a *Application) resolveBgBgAudioFiles(wordDir string) (string, string) {
+	return resolveAudioFileByBaseName(wordDir, "audio_front"), resolveAudioFileByBaseName(wordDir, "audio_back")
+}
+
+func (a *Application) hasAnyAudioFile(wordDir string) bool {
+	single := a.resolveSingleAudioFile(wordDir)
+	if single != "" {
+		return true
+	}
+
+	front, back := a.resolveBgBgAudioFiles(wordDir)
+	return front != "" || back != ""
+}
+
+func resolveAudioFileByBaseName(wordDir, baseName string) string {
+	entries, err := os.ReadDir(wordDir)
+	if err != nil {
+		return ""
+	}
+
+	prefix := baseName + "."
+	var resolved string
+	var resolvedModTime time.Time
+
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+
+		name := entry.Name()
+		if !strings.HasPrefix(name, prefix) {
+			continue
+		}
+
+		ext := strings.ToLower(filepath.Ext(name))
+		if _, ok := supportedAudioExtensions[ext]; !ok {
+			continue
+		}
+
+		info, err := entry.Info()
+		if err != nil {
+			continue
+		}
+
+		candidate := filepath.Join(wordDir, name)
+		if resolved == "" || info.ModTime().After(resolvedModTime) || (info.ModTime().Equal(resolvedModTime) && candidate < resolved) {
+			resolved = candidate
+			resolvedModTime = info.ModTime()
+		}
+	}
+
+	return resolved
+}
diff --git a/internal/gui/audio_paths_test.go b/internal/gui/audio_paths_test.go
new file mode 100644
index 0000000..1251c00
--- /dev/null
+++ b/internal/gui/audio_paths_test.go
@@ -0,0 +1,99 @@
+package gui
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+)
+
+func TestResolveSingleAudioFileFindsLegacyMp3WhenGuiDefaultIsWav(t *testing.T) {
+	tempDir := t.TempDir()
+	wordDir := filepath.Join(tempDir, "word")
+	if err := os.MkdirAll(wordDir, 0755); err != nil {
+		t.Fatalf("failed to create word dir: %v", err)
+	}
+
+	mp3Path := filepath.Join(wordDir, "audio.mp3")
+	if err := os.WriteFile(mp3Path, []byte("mp3"), 0644); err != nil {
+		t.Fatalf("failed to write mp3 file: %v", err)
+	}
+
+	app := &Application{
+		config: &Config{AudioFormat: "wav"},
+	}
+
+	got := app.resolveSingleAudioFile(wordDir)
+	if got != mp3Path {
+		t.Fatalf("resolveSingleAudioFile() = %q, want %q", got, mp3Path)
+	}
+}
+
+func TestResolveSingleAudioFilePrefersNewerOnDiskAudio(t *testing.T) {
+	tempDir := t.TempDir()
+	wordDir := filepath.Join(tempDir, "word")
+	if err := os.MkdirAll(wordDir, 0755); err != nil {
+		t.Fatalf("failed to create word dir: %v", err)
+	}
+
+	mp3Path := filepath.Join(wordDir, "audio.mp3")
+	wavPath := filepath.Join(wordDir, "audio.wav")
+	if err := os.WriteFile(mp3Path, []byte("mp3"), 0644); err != nil {
+		t.Fatalf("failed to write mp3 file: %v", err)
+	}
+	if err := os.WriteFile(wavPath, []byte("wav"), 0644); err != nil {
+		t.Fatalf("failed to write wav file: %v", err)
+	}
+
+	older := time.Now().Add(-time.Hour)
+	newer := time.Now()
+	if err := os.Chtimes(mp3Path, older, older); err != nil {
+		t.Fatalf("failed to set mp3 file time: %v", err)
+	}
+	if err := os.Chtimes(wavPath, newer, newer); err != nil {
+		t.Fatalf("failed to set wav file time: %v", err)
+	}
+
+	app := &Application{
+		config: &Config{AudioFormat: "wav"},
+	}
+
+	got := app.resolveSingleAudioFile(wordDir)
+	if got != wavPath {
+		t.Fatalf("resolveSingleAudioFile() = %q, want newer wav %q", got, wavPath)
+	}
+}
+
+func TestResolveBgBgAudioFilesFindLegacyMp3Files(t *testing.T) {
+	tempDir := t.TempDir()
+	wordDir := filepath.Join(tempDir, "word")
+	if err := os.MkdirAll(wordDir, 0755); err != nil {
+		t.Fatalf("failed to create word dir: %v", err)
+	}
+
+	frontPath := filepath.Join(wordDir, "audio_front.mp3")
+	backPath := filepath.Join(wordDir, "audio_back.mp3")
+	if err := os.WriteFile(frontPath, []byte("front"), 0644); err != nil {
+		t.Fatalf("failed to write front file: %v", err)
+	}
+	if err := os.WriteFile(backPath, []byte("back"), 0644); err != nil {
+		t.Fatalf("failed to write back file: %v", err)
+	}
+
+	older := time.Now().Add(-time.Hour)
+	if err := os.Chtimes(frontPath, older, older); err != nil {
+		t.Fatalf("failed to set front file time: %v", err)
+	}
+
+	app := &Application{
+		config: &Config{AudioFormat: "wav"},
+	}
+
+	gotFront, gotBack := app.resolveBgBgAudioFiles(wordDir)
+	if gotFront != frontPath {
+		t.Fatalf("resolveBgBgAudioFiles() front = %q, want %q", gotFront, frontPath)
+	}
+	if gotBack != backPath {
+		t.Fatalf("resolveBgBgAudioFiles() back = %q, want %q", gotBack, backPath)
+	}
+}
diff --git a/internal/gui/generator.go b/internal/gui/generator.go
index 6d1c883..1b9548a 100644
--- a/internal/gui/generator.go
+++ b/internal/gui/generator.go
@@ -62,11 +62,9 @@ func (a *Application) audioVoiceAndSpeed() (string, float64) {
 	switch a.audioProviderName() {
 	case "gemini":
 		if a.audioConfig != nil {
-			if voice := strings.TrimSpace(a.audioConfig.GeminiVoice); voice != "" {
-				return voice, a.geminiSpeed()
-			}
+			return strings.TrimSpace(a.audioConfig.GeminiVoice), a.geminiSpeed()
 		}
-		return randomVoice(a.audioVoices()), a.geminiSpeed()
+		return "", a.geminiSpeed()
 	default:
 		return randomVoice(a.audioVoices()), randomOpenAISpeed()
 	}
@@ -224,6 +222,10 @@ func (a *Application) generateAudioFront(ctx context.Context, word string, cardD
 	}
 	fmt.Printf("DEBUG (generateAudioFront): Successfully wrote front audio to: %s\n", frontFile)
 
+	if err := a.saveAudioAttribution(word, frontFile, voice, speed); err != nil {
+		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
+	}
+
 	// Update metadata
 	metadataFile := filepath.Join(cardDir, "audio_metadata.txt")
 	metadata := fmt.Sprintf("voice=%s\nspeed=%.2f\ncardtype=bg-bg\n", voice, speed)
@@ -261,6 +263,10 @@ func (a *Application) generateAudioBack(ctx context.Context, text string, cardDi
 	}
 	fmt.Printf("DEBUG (generateAudioBack): Successfully wrote back audio to: %s\n", backFile)
 
+	if err := a.saveAudioAttribution(text, backFile, voice, speed); err != nil {
+		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
+	}
+
 	return backFile, nil
 }
 
@@ -296,6 +302,9 @@ func (a *Application) generateAudioBgBg(ctx context.Context, front, back, cardDi
 	if err := a.saveAudioAttribution(front, frontFile, voice, speed); err != nil {
 		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
 	}
+	if err := a.saveAudioAttribution(back, backFile, voice, speed); err != nil {
+		fmt.Printf("Warning: Failed to save audio attribution: %v\n", err)
+	}
 
 	// Save voice metadata
 	metadataFile := filepath.Join(cardDir, "audio_metadata.txt")
diff --git a/internal/gui/generator_test.go b/internal/gui/generator_test.go
index 635b990..be0c254 100644
--- a/internal/gui/generator_test.go
+++ b/internal/gui/generator_test.go
@@ -220,7 +220,7 @@ func TestGenerateAudioUsesSharedOpenAIVoices(t *testing.T) {
 	}
 }
 
-func TestGenerateAudioUsesSharedGeminiVoicesAndAttribution(t *testing.T) {
+func TestGenerateAudioUsesGeminiModelDefaultVoiceAndAttribution(t *testing.T) {
 	originalFactory := newAudioProvider
 	t.Cleanup(func() {
 		newAudioProvider = originalFactory
@@ -257,6 +257,7 @@ func TestGenerateAudioUsesSharedGeminiVoicesAndAttribution(t *testing.T) {
 			OutputDir:      tempDir,
 			GoogleAPIKey:   "google-key",
 			GeminiTTSModel: "gemini-2.5-flash-preview-tts",
+			GeminiVoice:    "",
 		},
 	}
 
@@ -271,8 +272,8 @@ func TestGenerateAudioUsesSharedGeminiVoicesAndAttribution(t *testing.T) {
 	if capturedConfig.Provider != "gemini" {
 		t.Fatalf("captured Provider = %q, want %q", capturedConfig.Provider, "gemini")
 	}
-	if capturedConfig.GeminiVoice != "sentinel-gemini-voice" {
-		t.Fatalf("captured GeminiVoice = %q, want %q", capturedConfig.GeminiVoice, "sentinel-gemini-voice")
+	if capturedConfig.GeminiVoice != "" {
+		t.Fatalf("captured GeminiVoice = %q, want empty model-default voice", capturedConfig.GeminiVoice)
 	}
 	if capturedConfig.OutputFormat != "wav" {
 		t.Fatalf("captured OutputFormat = %q, want %q", capturedConfig.OutputFormat, "wav")
@@ -293,8 +294,8 @@ func TestGenerateAudioUsesSharedGeminiVoicesAndAttribution(t *testing.T) {
 	if !strings.Contains(attribution, "Audio generated by Google Gemini TTS") {
 		t.Fatalf("gemini attribution missing header: %q", attribution)
 	}
-	if !strings.Contains(attribution, "Voice: sentinel-gemini-voice") {
-		t.Fatalf("gemini attribution missing voice: %q", attribution)
+	if strings.Contains(attribution, "sentinel-gemini-voice") {
+		t.Fatalf("gemini attribution should not use the shared voice list when voice is unset: %q", attribution)
 	}
 }
 
@@ -432,6 +433,19 @@ func TestGenerateAudioFrontUsesSharedOpenAIVoices(t *testing.T) {
 	if !strings.HasSuffix(outputPath, "audio_front.mp3") {
 		t.Fatalf("outputPath = %q, want audio_front.mp3", outputPath)
 	}
+
+	attrPath := audio.AttributionPath(outputPath)
+	attributionData, err := os.ReadFile(attrPath)
+	if err != nil {
+		t.Fatalf("expected attribution file %q: %v", attrPath, err)
+	}
+	attribution := string(attributionData)
+	if !strings.Contains(attribution, "Audio generated by OpenAI TTS") {
+		t.Fatalf("front attribution missing header: %q", attribution)
+	}
+	if !strings.Contains(attribution, "Voice: sentinel-front-voice") {
+		t.Fatalf("front attribution missing voice: %q", attribution)
+	}
 }
 
 func TestGenerateAudioBackUsesSharedOpenAIVoices(t *testing.T) {
@@ -497,6 +511,19 @@ func TestGenerateAudioBackUsesSharedOpenAIVoices(t *testing.T) {
 	if !strings.HasSuffix(outputPath, "audio_back.mp3") {
 		t.Fatalf("outputPath = %q, want audio_back.mp3", outputPath)
 	}
+
+	attrPath := audio.AttributionPath(outputPath)
+	attributionData, err := os.ReadFile(attrPath)
+	if err != nil {
+		t.Fatalf("expected attribution file %q: %v", attrPath, err)
+	}
+	attribution := string(attributionData)
+	if !strings.Contains(attribution, "Audio generated by OpenAI TTS") {
+		t.Fatalf("back attribution missing header: %q", attribution)
+	}
+	if !strings.Contains(attribution, "Bulgarian word: круша") {
+		t.Fatalf("back attribution missing spoken text: %q", attribution)
+	}
 }
 
 func TestGenerateAudioProviderFactoryError(t *testing.T) {
diff --git a/internal/gui/navigation.go b/internal/gui/navigation.go
index 59b5492..cc1d9e6 100644
--- a/internal/gui/navigation.go
+++ b/internal/gui/navigation.go
@@ -95,22 +95,10 @@ func (a *Application) scanExistingWords() {
 		hasContent := false
 
 		// Check for audio file (both en-bg and bg-bg formats)
-		audioFile := filepath.Join(wordDir, fmt.Sprintf("audio.%s", a.config.AudioFormat))
-		if _, err := os.Stat(audioFile); err == nil {
+		if a.hasAnyAudioFile(wordDir) {
 			hasContent = true
 		}
 
-		// Check for bg-bg audio files (audio_front and audio_back)
-		if !hasContent {
-			frontAudio := filepath.Join(wordDir, fmt.Sprintf("audio_front.%s", a.config.AudioFormat))
-			backAudio := filepath.Join(wordDir, fmt.Sprintf("audio_back.%s", a.config.AudioFormat))
-			if _, errFront := os.Stat(frontAudio); errFront == nil {
-				hasContent = true
-			} else if _, errBack := os.Stat(backAudio); errBack == nil {
-				hasContent = true
-			}
-		}
-
 		// Check for image files
 		if !hasContent {
 			patterns := []string{
@@ -310,7 +298,9 @@ func (a *Application) loadWordByIndex(index int) {
 			// Load from queue job
 			a.currentTranslation = job.Translation
 			a.currentAudioFile = job.AudioFile
+			a.currentAudioFileBack = job.AudioFileBack
 			a.currentImage = job.ImageFile
+			a.currentCardType = job.CardType
 
 			fyne.Do(func() {
 				if job.Translation != "" {
@@ -319,6 +309,9 @@ func (a *Application) loadWordByIndex(index int) {
 				if job.AudioFile != "" {
 					a.audioPlayer.SetAudioFile(job.AudioFile)
 				}
+				if job.AudioFileBack != "" {
+					a.audioPlayer.SetBackAudioFile(job.AudioFileBack)
+				}
 				if job.ImageFile != "" {
 					a.imageDisplay.SetImages([]string{job.ImageFile})
 				}
@@ -446,10 +439,9 @@ func (a *Application) loadExistingFiles(word string) {
 	if cardType.IsBgBg() {
 		fmt.Printf("DEBUG (loadExistingFiles): Loading audio files for bg-bg card\n")
 		// For bg-bg cards, load both front and back audio
-		frontAudio := filepath.Join(wordDir, fmt.Sprintf("audio_front.%s", a.config.AudioFormat))
-		backAudio := filepath.Join(wordDir, fmt.Sprintf("audio_back.%s", a.config.AudioFormat))
+		frontAudio, backAudio := a.resolveBgBgAudioFiles(wordDir)
 
-		if _, err := os.Stat(frontAudio); err == nil {
+		if frontAudio != "" {
 			a.currentAudioFile = frontAudio
 			fmt.Printf("DEBUG (loadExistingFiles): Found front audio: %s\n", frontAudio)
 			fyne.Do(func() {
@@ -459,7 +451,7 @@ func (a *Application) loadExistingFiles(word string) {
 			fmt.Printf("DEBUG (loadExistingFiles): Front audio not found: %s\n", frontAudio)
 		}
 
-		if _, err := os.Stat(backAudio); err == nil {
+		if backAudio != "" {
 			a.currentAudioFileBack = backAudio
 			fmt.Printf("DEBUG (loadExistingFiles): Found back audio: %s\n", backAudio)
 			fyne.Do(func() {
@@ -471,8 +463,8 @@ func (a *Application) loadExistingFiles(word string) {
 	} else {
 		fmt.Printf("DEBUG (loadExistingFiles): Loading audio files for en-bg card\n")
 		// For en-bg cards, load standard audio file
-		audioFile := filepath.Join(wordDir, fmt.Sprintf("audio.%s", a.config.AudioFormat))
-		if _, err := os.Stat(audioFile); err == nil {
+		audioFile := a.resolveSingleAudioFile(wordDir)
+		if audioFile != "" {
 			a.currentAudioFile = audioFile
 			fmt.Printf("DEBUG (loadExistingFiles): Found audio: %s\n", audioFile)
 			fyne.Do(func() {
@@ -581,31 +573,31 @@ func (a *Application) checkForMissingFiles(word string) {
 
 	// Check for missing audio file
 	if a.currentAudioFile == "" {
-		// First check for en-bg audio file
-		audioFile := filepath.Join(wordDir, fmt.Sprintf("audio.%s", a.config.AudioFormat))
-		if _, err := os.Stat(audioFile); err == nil {
-			a.currentAudioFile = audioFile
-			fyne.Do(func() {
-				a.audioPlayer.SetAudioFile(audioFile)
-				a.updateStatus(fmt.Sprintf("Found audio file for %s", word))
-			})
-		} else {
-			// Check for bg-bg audio_front file
-			frontAudio := filepath.Join(wordDir, fmt.Sprintf("audio_front.%s", a.config.AudioFormat))
-			if _, err := os.Stat(frontAudio); err == nil {
+		if a.currentCardType == "bg-bg" {
+			frontAudio, _ := a.resolveBgBgAudioFiles(wordDir)
+			if frontAudio != "" {
 				a.currentAudioFile = frontAudio
 				fyne.Do(func() {
 					a.audioPlayer.SetAudioFile(frontAudio)
 					a.updateStatus(fmt.Sprintf("Found audio file for %s", word))
 				})
 			}
+		} else {
+			audioFile := a.resolveSingleAudioFile(wordDir)
+			if audioFile != "" {
+				a.currentAudioFile = audioFile
+				fyne.Do(func() {
+					a.audioPlayer.SetAudioFile(audioFile)
+					a.updateStatus(fmt.Sprintf("Found audio file for %s", word))
+				})
+			}
 		}
 	}
 
 	// Check for missing back audio file (bg-bg cards)
 	if a.currentAudioFileBack == "" {
-		backAudio := filepath.Join(wordDir, fmt.Sprintf("audio_back.%s", a.config.AudioFormat))
-		if _, err := os.Stat(backAudio); err == nil {
+		_, backAudio := a.resolveBgBgAudioFiles(wordDir)
+		if backAudio != "" {
 			a.currentAudioFileBack = backAudio
 			fyne.Do(func() {
 				a.audioPlayer.SetBackAudioFile(backAudio)