dialogues = [
    {"character": "Mochi", "text": "Hi there!", "emotion": "cheerful"},
    {"character": "Hero", "text": "Hey Mochi!", "emotion": "excited"},
]

@dataclass
class TTSResult:
    audio_path: Path
    duration_seconds: float
    voice_used: str
    character_name: str

class TTSGenerator:
    async def generate_dialogue(
        self,
        dialogue: str,
        character_name: str,
        voice_name: str = None,
        emotion: str = "neutral",
    ) -> TTSResult:
        """
        Generate speech for a single character's dialogue.

        Args:
            dialogue: Text to speak
            character_name: Who is speaking
            voice_name: Specific voice (or auto-select)
            emotion: Emotional tone hint

        Returns:
            TTSResult with audio path and duration
        """
        ...

    async def generate_conversation(
        self,
        dialogues: list[dict],
        voice_mapping: dict[str, str] = None,
    ) -> list[TTSResult]:
        """
        Generate speech for a multi-character conversation.

        Maintains voice consistency across the conversation.
        Max 2 speakers per call (Gemini TTS limit).

        Args:
            dialogues: List of {character, text, emotion} dicts
            voice_mapping: Optional {character_name: voice_name} mapping

        Returns:
            List of TTSResult, one per dialogue line
        """
        ...

    async def generate_panel_audio(
        self,
        panel_dialogues: list[str],
        character_names: list[str],
        voice_mapping: dict[str, str] = None,
    ) -> tuple[Path, float]:
        """
        Generate combined audio for a single manga panel.

        Multiple characters speaking in one panel get combined
        into a single audio file.

        Returns:
            (combined_audio_path, total_duration_seconds)
        """
        ...

from skills.generate_tts import TTSGenerator

tts = TTSGenerator()

# Single dialogue
result = await tts.generate_dialogue(
    dialogue="Let's go on an adventure!",
    character_name="Mochi",
    emotion="excited"
)
print(f"Audio: {result.audio_path}, Duration: {result.duration_seconds}s")

# Multi-character conversation
results = await tts.generate_conversation(
    dialogues=[
        {"character": "Mochi", "text": "What's that?", "emotion": "curious"},
        {"character": "Hero", "text": "It's a treasure map!", "emotion": "excited"},
    ],
    voice_mapping={
        "Mochi": "Leda",   # Youthful voice for pet
        "Hero": "Kore",    # Firm voice for protagonist
    }
)

Panel Dialogue → TTS → duration_seconds → Veo (duration param)
                  ↓
             audio_path → FFmpeg (combine with video)