PocketTTS

import FluidAudioTTS

let manager = PocketTtsManager()
try await manager.initialize()

let audioData = try await manager.synthesize(text: "Hello, world!")

try await manager.synthesizeToFile(
    text: "Hello, world!",
    outputURL: URL(fileURLWithPath: "/tmp/output.wav")
)

PocketTtsManager.synthesize(text:)
  → chunkText() — split into max 50 token chunks
  → loadMimiInitialState() — 23 streaming state tensors
  → FOR EACH CHUNK:
      → tokenizer.encode() — SentencePiece
      → embedTokens() — table lookup
      → prefillKVCache() — 125 voice + N text tokens
      → GENERATE LOOP:
          → runFlowLMStep() — transformer_out + eos_logit
          → flowDecode() — 8 Euler steps → 32-dim latent
          → denormalize() → quantize() → runMimiDecoder()
          → 1920 audio samples per frame
  → concatenate + postprocess
  → WAV output (24kHz mono)

text → SentencePiece tokenizer → subword tokens → PocketTTS model → audio
                                                    ↑
                                          pronunciation decisions
                                          happen inside model weights
                                          (no external control)