refactor: add app directory, dir structure change

41 files changed, 123 insertions, 106 deletions

diff --git a/.gitignore b/.gitignore
index ca8252f..313ac60 100644
--- a/.gitignore
+++ b/.gitignore
@@ -10,12 +10,12 @@ voices/*
 !voices/.gitkeep
 
 # Backend checkouts cloned by the setup wizards (backends.audiocpp / .faster)
-/audio.cpp/
-/faster-qwen3-tts/
+/app/audio.cpp/
+/app/faster-qwen3-tts/
 
 # Managed Python environment created by audiobook.py (envs.py) for the app
 # requirements and the backend TTS packages (qwen-tts, faster-qwen3-tts).
-/envs/
+/app/envs/
 
 *.epub
 input/*.txt
diff --git a/README.md b/README.md
index 3f61655..5d96ff8 100644
--- a/README.md
+++ b/README.md
@@ -21,7 +21,7 @@ The converter sends text extracted from your books to a locally running TTS serv
 ## Installation
 
 Clone the repo. No manual environment setup is needed — `audiobook.py`
-creates and manages its own virtual environment (`envs/tts`) the first time
+creates and manages its own virtual environment (`app/envs/tts`) the first time
 it runs, installing its requirements and any backend TTS packages into it.
 
 ```bash
@@ -30,7 +30,7 @@ cd tts-audiobook-generator
 python audiobook.py
 ```
 
-On first launch `audiobook.py` creates `envs/tts` (via `python -m venv`),
+On first launch `audiobook.py` creates `app/envs/tts` (via `python -m venv`),
 installs `requirements.txt` into it, and re-launches itself inside that
 environment. Backend packages (`qwen-tts`, `faster-qwen3-tts[demo]`) are
 pip-installed into the same venv by their setup wizards.
@@ -39,7 +39,7 @@ To add extras manually (e.g. FlashAttention), pip-install into the managed
 venv directly:
 
 ```bash
-envs/tts/bin/python -m pip install flash-attn
+app/envs/tts/bin/python -m pip install flash-attn
 ```
 
 Put your book files (epub, etc.) in the `input/` directory. The output goes to `output/`.
@@ -61,7 +61,7 @@ A full-screen TUI opens and shows each backend's status in a table — **unavail
 
 **Configure a backend…** and **Server…** only appear once at least one backend is installed or running.
 
-Everything the TUI does can also be scripted with flags: `python audiobook.py --backend audiocpp --model higgs --voice narrator`, or `python -m backends.audiocpp --families higgs_audio_tts --clone --build-backend cuda`.
+Everything the TUI does can also be scripted with flags: `python audiobook.py --backend audiocpp --model higgs --voice narrator`, or `python app/backends/audiocpp.py --families higgs_audio_tts --clone --build-backend cuda`.
 
 You need one of the following backends (the TUI sets them up for you; manual steps below):
 
@@ -90,15 +90,15 @@ You need one of the following backends (the TUI sets them up for you; manual ste
 | `--transcription "..."`            | `qwen`: Override whisper auto-transcription with manual audio transcript.                                                                                                                                                                                          |
 | `--no-transcription`               | `qwen`: Skip auto-transcription of the reference audio.                                                                                                                                                                                                            |
 
-Other options including backend server URLs/ports are configured in `converter/config.py`
+Other options including backend server URLs/ports are configured in `app/converter/config.py`
 
 ## TTS Backend Setup
 
-Installation and usage documentation for each supported TTS backend is in the `docs/` directory:
+Installation and usage documentation for each supported TTS backend is in the `app/docs/` directory:
 
-- [audio.cpp instructions](docs/backend-audiocpp.md)
-- [qwen-tts instructions](docs/backend-qwen.md)
-- [faster-qwen-tts instructions](docs/backend-faster.md)
+- [audio.cpp instructions](app/docs/backend-audiocpp.md)
+- [qwen-tts instructions](app/docs/backend-qwen.md)
+- [faster-qwen-tts instructions](app/docs/backend-faster.md)
 
 ## Tips
 
diff --git a/backends/__init__.py b/app/backends/__init__.py
index 6a23ab7..ed772d4 100644
--- a/backends/__init__.py
+++ b/app/backends/__init__.py
@@ -47,7 +47,7 @@ class BackendStatus:
 
     INSTALLED means the backend itself is present (a cloned + built
     checkout, or a pip package). CONFIGURED means the supporting files are
-    in place (a server.json / voices.json and a converter/config.py that
+    in place (a server.json / voices.json and an app/converter/config.py that
     points at the right port). RUNNING means an external server is
     currently accepting connections on the configured port (probed by
     ``backends.common.server_running``). DETAILS are short status lines for
diff --git a/backends/audiocpp.py b/app/backends/audiocpp.py
index 486017f..cc67efc 100755
--- a/backends/audiocpp.py
+++ b/app/backends/audiocpp.py
@@ -2,10 +2,10 @@
 """Set up the audio.cpp TTS backend for the audiobook generator.
 
 This does the whole audio.cpp setup end-to-end as a full-screen DOS-style
-TUI: locate or clone an audio.cpp checkout into ``./audio.cpp``, optionally
+TUI: locate or clone an audio.cpp checkout into ``app/audio.cpp``, optionally
 build ``audiocpp_server``, pick model families/packages from the checkout's
 ``model_specs`` catalog, transcribe reference .wav voices, write
-``server.json``, sync ``converter/config.py``, download the models, and
+``server.json``, sync ``app/converter/config.py``, download the models, and
 print the exact command to start the server. It is driven by
 ``audiobook.py``'s TUI hub (``backends.REGISTRY``) but can also be run
 directly for scripting — every value has a flag, and a non-interactive run
@@ -16,7 +16,7 @@ entry from ``GET /v1/models`` at startup), so any TTS family listed in the
 catalog works without further changes.
 
 Usage:
-    python -m backends.audiocpp [--wavs WAV_DIR] [--output PATH]
+    python app/backends/audiocpp.py [--wavs WAV_DIR] [--output PATH]
         [--audiocpp-dir PATH] [--clone] [--families FAM1,FAM2]
         [--all-packages] [--host HOST] [--port PORT]
         [--build-backend {cuda,vulkan,hip,cpu}] [--backend {cuda,vulkan,hip,cpu}]
@@ -38,7 +38,7 @@ import urllib.parse
 from pathlib import Path
 from typing import Callable, Dict, List, Optional, Set, Tuple
 
-# Allow running directly (python backends/audiocpp.py) from any cwd.
+# Allow running directly (python app/backends/audiocpp.py) from any cwd.
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
 
 from backends import (
@@ -49,6 +49,7 @@ from backends import (
     format_launch_hint,
 )
 from backends.common import (
+    APP_DIR,
     CONFIG_PATH,
     PROMPT_TEXT_FILENAME,
     TTS_ROOT,
@@ -213,7 +214,7 @@ def _backend_options(detected: Optional[str] = None
 
 
 def config_port() -> int:
-    """Return the port of AUDIOCPP_API_URL in converter/config.py."""
+    """Return the port of AUDIOCPP_API_URL in app/converter/config.py."""
     try:
         return urllib.parse.urlsplit(config.AUDIOCPP_API_URL).port or FALLBACK_PORT
     except ValueError:
@@ -228,7 +229,7 @@ def _url_with_port(url: str, port: int) -> str:
 
 
 def update_config_api_url_port(port: int, config_path: Optional[Path] = None) -> bool:
-    """Rewrite the port inside AUDIOCPP_API_URL in converter/config.py.
+    """Rewrite the port inside AUDIOCPP_API_URL in app/converter/config.py.
 
     Only the quoted URL literal is replaced; surrounding lines and the
     trailing comment are preserved. Returns True when the file was changed.
@@ -296,14 +297,16 @@ def default_model_id(family: str) -> str:
 def detect_audiocpp_dir() -> Optional[Path]:
     """Best-effort location of a local audio.cpp checkout with model_specs.
 
-    Checks the AUDIOCPP_DIR environment variable, then an ``audio.cpp``
-    directory in or above the current working directory. Returns the path
-    only when it contains a ``model_specs`` directory.
+    Checks the AUDIOCPP_DIR environment variable, then ``app/audio.cpp`` in
+    the tts-audiobook-generator root, then an ``audio.cpp`` directory in or
+    above the current working directory. Returns the path only when it
+    contains a ``model_specs`` directory.
     """
     candidates: List[Path] = []
     env_dir = os.environ.get("AUDIOCPP_DIR")
     if env_dir:
         candidates.append(Path(os.path.expanduser(env_dir)))
+    candidates.append(APP_DIR / AUDIOCPP_DIR_NAME)
     cwd = Path.cwd()
     candidates.append(cwd / "audio.cpp")
     candidates.append(cwd.parent / "audio.cpp")
@@ -553,7 +556,7 @@ def print_empty_transcript_warning(transcripts: Dict[str, str]) -> None:
 
 
 def _apply_port_sync(port: int, accepted: bool) -> None:
-    """Write the port into converter/config.py, or report when declined."""
+    """Write the port into app/converter/config.py, or report when declined."""
     if accepted:
         if not update_config_api_url_port(port):
             print(f"[WARNING] Could not update {CONFIG_PATH}; edit "
@@ -657,7 +660,7 @@ def _flag_plan(wav_files: list, prompt_path: Path, force: bool) -> dict:
 
 
 def _offer_config_model_id_sync(model_id: str, accepted: Optional[bool]) -> None:
-    """Point converter/config.py at a single hosted model entry.
+    """Point app/converter/config.py at a single hosted model entry.
 
     The converter requests the model id configured in AUDIOCPP_MODEL_ID,
     and single-model servers use the same id for the clone entry, so both
@@ -905,20 +908,21 @@ def _wizard(stdscr, args: argparse.Namespace, parser: argparse.ArgumentParser
                         audiocpp_dir = do_browse()
                     else:
                         # No checkout found anywhere: offer to clone one into
-                        # ./audio.cpp or browse for an existing checkout. Esc
-                        # on this first menu aborts the wizard.
+                        # ./app/audio.cpp or browse for an existing checkout.
+                        # Esc on this first menu aborts the wizard.
                         choice = tui.menu(
                             stdscr, "No audio.cpp checkout found",
-                            [(f"Clone into ./{AUDIOCPP_DIR_NAME} "
+                            [(f"Clone into ./app/{AUDIOCPP_DIR_NAME} "
                               f"(from {AUDIOCPP_GIT_URL})", "clone"),
                              ("Browse for an existing checkout", "browse")],
                             help_lines=[
                                 "audio.cpp hosts the TTS model families "
                                 "this generator uses.",
-                                "Clone it into the project directory, or "
-                                "point at an existing checkout."])
+                                "Clone it into the project's app "
+                                "directory, or point at an existing "
+                                "checkout."])
                         if choice == "clone":
-                            target = TTS_ROOT / AUDIOCPP_DIR_NAME
+                            target = APP_DIR / AUDIOCPP_DIR_NAME
                             with tui.suspend(stdscr):
                                 rc = common.git_clone(AUDIOCPP_GIT_URL,
                                                       target)
@@ -1098,7 +1102,7 @@ def _wizard(stdscr, args: argparse.Namespace, parser: argparse.ArgumentParser
             sync_port: Optional[bool] = None
             if port != config_port():
                 sync_port = tui.confirm(
-                    stdscr, f"Update AUDIOCPP_API_URL in converter/config.py "
+                    stdscr, f"Update AUDIOCPP_API_URL in app/converter/config.py "
                     f"to port {port} so audiobook.py talks to this server",
                     default=True, cancel_value=_GO_BACK)
                 if sync_port is _GO_BACK:
@@ -1194,7 +1198,7 @@ def _wizard(stdscr, args: argparse.Namespace, parser: argparse.ArgumentParser
                     and config.AUDIOCPP_CLONE_MODEL_ID == entry_ids[0]):
                 sync_model_ids = tui.confirm(
                     stdscr, "Update AUDIOCPP_MODEL_ID and "
-                    "AUDIOCPP_CLONE_MODEL_ID in converter/config.py to "
+                    "AUDIOCPP_CLONE_MODEL_ID in app/converter/config.py to "
                     f"'{entry_ids[0]}' so audiobook.py uses this model",
                     default=True, cancel_value=_GO_BACK)
                 if sync_model_ids is _GO_BACK:
@@ -1239,16 +1243,16 @@ def _wizard(stdscr, args: argparse.Namespace, parser: argparse.ArgumentParser
 def find_local_checkout() -> Optional[Path]:
     """Best-effort location of an audio.cpp checkout with model_specs.
 
-    Checks the AUDIOCPP_DIR environment variable, then ``./audio.cpp`` inside
-    the tts-audiobook-generator root, then an ``audio.cpp`` directory in or
-    above the current working directory. Returns the path only when it
+    Checks the AUDIOCPP_DIR environment variable, then ``app/audio.cpp``
+    inside the tts-audiobook-generator root, then an ``audio.cpp`` directory
+    in or above the current working directory. Returns the path only when it
     contains a ``model_specs`` directory.
     """
     candidates: List[Path] = []
     env_dir = os.environ.get("AUDIOCPP_DIR")
     if env_dir:
         candidates.append(Path(os.path.expanduser(env_dir)))
-    candidates.append(TTS_ROOT / AUDIOCPP_DIR_NAME)
+    candidates.append(APP_DIR / AUDIOCPP_DIR_NAME)
     cwd = Path.cwd()
     candidates.append(cwd / AUDIOCPP_DIR_NAME)
     candidates.append(cwd.parent / AUDIOCPP_DIR_NAME)
@@ -1446,7 +1450,7 @@ def _collect_from_flags(args: argparse.Namespace,
     if audiocpp_dir is None:
         audiocpp_dir = find_local_checkout()
     if audiocpp_dir is None and args.clone:
-        target = TTS_ROOT / AUDIOCPP_DIR_NAME
+        target = APP_DIR / AUDIOCPP_DIR_NAME
         rc = common.git_clone(AUDIOCPP_GIT_URL, target)
         if rc != 0:
             parser.error(f"git clone failed (exit {rc}); clone audio.cpp "
@@ -1455,7 +1459,7 @@ def _collect_from_flags(args: argparse.Namespace,
     if audiocpp_dir is None:
         parser.error(
             "An audio.cpp checkout is required. Pass --audiocpp-dir PATH, "
-            "or --clone to clone ./audio.cpp, or run without flags for the "
+            "or --clone to clone app/audio.cpp, or run without flags for the "
             "TUI wizard.")
     audiocpp_dir = Path(audiocpp_dir).resolve()
     if not audiocpp_dir.is_dir():
@@ -1588,7 +1592,7 @@ def build_parser() -> argparse.ArgumentParser:
     """The audio.cpp setup CLI (also used to build a default namespace)."""
     parser = argparse.ArgumentParser(
         description="Set up the audio.cpp TTS backend: clone/build, pick "
-                    "models, write server.json, and sync converter/config.py.")
+                    "models, write server.json, and sync app/converter/config.py.")
     parser.add_argument("--wavs", type=resolve_wav_dir_arg, default=None,
                         dest="input_dir", metavar="WAV_DIR",
                         help="Directory with .wav reference files to publish as "
@@ -1603,11 +1607,11 @@ def build_parser() -> argparse.ArgumentParser:
     parser.add_argument("--audiocpp-dir", type=normalize_dir_arg, default=None,
                         help="Path to a local audio.cpp checkout containing a "
                              "model_specs/ directory (default: detected from "
-                             "AUDIOCPP_DIR or ./audio.cpp; in the TUI you can "
+                             "AUDIOCPP_DIR or ./app/audio.cpp; in the TUI you can "
                              "clone one instead)")
     parser.add_argument("--clone", action="store_true",
                         help="Non-interactive: clone audio.cpp into "
-                             "./audio.cpp when no checkout is found")
+                             "./app/audio.cpp when no checkout is found")
     parser.add_argument("--families", type=str, default=None,
                         help="Comma-separated model families to host, as named "
                              "in the audio.cpp catalog (e.g. "
@@ -1623,7 +1627,7 @@ def build_parser() -> argparse.ArgumentParser:
                         help="Bind host for the server (default: 127.0.0.1)")
     parser.add_argument("--port", type=int, default=None,
                         help="Port for the server (default: the port in "
-                             "AUDIOCPP_API_URL from converter/config.py)")
+                             "AUDIOCPP_API_URL from app/converter/config.py)")
     parser.add_argument("--backend", choices=BACKENDS, default=None,
                         help="Inference backend recorded in server.json "
                              "(default: auto-detected from the checkout's "
@@ -1646,7 +1650,7 @@ def build_parser() -> argparse.ArgumentParser:
                              "only)")
     parser.add_argument("--no-sync-port", action="store_true",
                         help="Do not rewrite AUDIOCPP_API_URL in "
-                             "converter/config.py when --port differs")
+                             "app/converter/config.py when --port differs")
     parser.add_argument("--no-sync-model-ids", action="store_true",
                         help="Do not rewrite AUDIOCPP_MODEL_ID/"
                              "AUDIOCPP_CLONE_MODEL_ID for a single-entry server")
@@ -1665,7 +1669,7 @@ def detect() -> BackendStatus:
         return BackendStatus("audiocpp", "audio.cpp", installed=False,
                              configured=False, running=running,
                              details=["not cloned — run setup to clone "
-                                      "./audio.cpp"])
+                                      "./app/audio.cpp"])
     details.append(f"checkout: {checkout}")
     binary = find_audiocpp_server_bin(checkout)
     built = binary is not None
diff --git a/backends/common.py b/app/backends/common.py
index 2529a8f..42faa7e 100644
--- a/backends/common.py
+++ b/app/backends/common.py
@@ -3,7 +3,7 @@
 Every TTS backend setup wizard (audio.cpp, qwen, faster) lives in its own
 module under ``backends``; this module holds the pieces more than one of
 them needs: .wav discovery, path normalization, and the regex edit that
-keeps ``converter/config.py`` in sync with the choices made in a wizard.
+keeps ``app/converter/config.py`` in sync with the choices made in a wizard.
 It deliberately imports nothing from the other backend modules (or the
 TUI) so it can be reused without pulling curses into a non-interactive
 run.
@@ -16,18 +16,22 @@ from pathlib import Path
 from typing import Dict, List, Optional, Set, Tuple
 
 # The tts-audiobook-generator checkout root (where audiobook.py lives).
-# Backend checkouts are cloned into subdirectories of this root
-# (./audio.cpp, ./faster-qwen3-tts) so a single tree holds everything.
-TTS_ROOT = Path(__file__).resolve().parent.parent
+# Everything non-user-facing lives under ./app: the source packages
+# (backends, converter, ui), the generated dirs (envs, chunks, logs, debug),
+# and the backend checkouts (app/audio.cpp, app/faster-qwen3-tts).
+TTS_ROOT = Path(__file__).resolve().parent.parent.parent
+
+# The single "everything else" directory under TTS_ROOT.
+APP_DIR = TTS_ROOT / "app"
 
 # The project's sample-voice directory: .wav files dropped here are offered
 # as the default source when a setup/configure wizard asks for a wav
 # directory (both the TUI browser start and the --wavs flag default).
 VOICES_DIR = TTS_ROOT / "voices"
 
-# converter/config.py — rewritten in place by update_config_value so the
+# app/converter/config.py — rewritten in place by update_config_value so the
 # converter picks up the host/port/voice a wizard configured.
-CONFIG_PATH = TTS_ROOT / "converter" / "config.py"
+CONFIG_PATH = APP_DIR / "converter" / "config.py"
 
 # Output directory of tts-audiobook-generator; never offered as a .wav
 # source by detect_wav_dir.
@@ -173,7 +177,7 @@ def server_running(url: str, timeout: float = 0.3) -> bool:
 
 def update_config_value(key: str, value: str,
                         config_path: Optional[Path] = None) -> bool:
-    """Rewrite a ``KEY = "value"`` line in converter/config.py.
+    """Rewrite a ``KEY = "value"`` line in app/converter/config.py.
 
     Only the quoted literal is replaced; surrounding lines and the trailing
     comment are preserved. Returns True when the file was changed. Used by
diff --git a/backends/envs.py b/app/backends/envs.py
index 7596db6..6e5b6cc 100644
--- a/backends/envs.py
+++ b/app/backends/envs.py
@@ -2,7 +2,7 @@
 
 audiobook.py is meant to be launched from any Python (a bare system interpreter
 is fine): on startup it bootstraps a single tool-managed venv at
-``envs/tts`` and re-execs itself inside it. That venv holds both the
+``app/envs/tts`` and re-execs itself inside it. That venv holds both the
 audiobook app's own ``requirements.txt`` dependencies and the backend TTS
 packages (``qwen-tts``, ``faster-qwen3-tts[demo]``) the setup wizards pip
 install, so nothing is ever installed into the launching interpreter's
@@ -11,7 +11,7 @@ environment.
 A parent process never needs to "activate" an environment — activation is
 just a shell convenience that puts an env's ``bin`` on PATH. Instead every
 helper here resolves the env's binaries by absolute path
-(``envs/tts/bin/python``, ``envs/tts/bin/qwen-tts-demo``), so the hub can
+(``app/envs/tts/bin/python``, ``app/envs/tts/bin/qwen-tts-demo``), so the hub can
 spawn servers in this env from any parent environment.
 
 This module is imported before audiobook.py's third-party dependencies, so
@@ -28,10 +28,10 @@ from typing import List
 from backends import common
 
 # The tts-audiobook-generator checkout root (where audiobook.py lives).
-TTS_ROOT = Path(__file__).resolve().parent.parent
+TTS_ROOT = Path(__file__).resolve().parent.parent.parent
 
 # One shared venv for the app requirements and every pip-installed backend.
-ENV_DIR = TTS_ROOT / "envs" / "tts"
+ENV_DIR = TTS_ROOT / "app" / "envs" / "tts"
 REQUIREMENTS_PATH = TTS_ROOT / "requirements.txt"
 
 # Marker file recording the requirements.txt hash last installed into the env,
diff --git a/backends/faster.py b/app/backends/faster.py
index 50c6102..0d34a0f 100755
--- a/backends/faster.py
+++ b/app/backends/faster.py
@@ -7,11 +7,11 @@ reference voice configured on the server through a ``voices.json``. This
 module sets the whole backend up end-to-end as a TUI: pip-install the
 package, clone the repo (for ``examples/openai_server.py``), build a
 ``voices.json`` from a directory of .wav references (transcribed with
-Whisper), sync ``converter/config.py``, and print the launch command. It is
+Whisper), sync ``app/converter/config.py``, and print the launch command. It is
 driven by ``audiobook.py``'s hub but can also be run directly with flags.
 
 Usage:
-    python -m backends.faster [--wavs WAV_DIR] [--output PATH]
+    python app/backends/faster.py [--wavs WAV_DIR] [--output PATH]
         [--language LANG] [--whisper-model NAME] [--force]
         [--port PORT] [--voice NAME] [--skip-install] [--skip-clone]
 """
@@ -33,7 +33,7 @@ from backends import (
     format_launch_hint,
 )
 from backends.common import (
-    TTS_ROOT,
+    APP_DIR,
     VOICES_DIR,
     find_wav_files,
     normalize_dir_arg,
@@ -53,7 +53,7 @@ WHISPER_MODELS = ("tiny", "base", "small", "medium", "large-v3")
 
 
 def _checkout() -> Path:
-    return TTS_ROOT / FASTER_DIR_NAME
+    return APP_DIR / FASTER_DIR_NAME
 
 
 def _is_installed() -> bool:
@@ -135,7 +135,7 @@ def _wizard(stdscr, args: argparse.Namespace) -> Optional[dict]:
     do_clone = False
     if not _is_cloned() and not args.skip_clone:
         choice = confirm(f"faster-qwen3-tts repo not cloned. Clone it into "
-                         f"./{FASTER_DIR_NAME}?", default=True)
+                         f"./app/{FASTER_DIR_NAME}?", default=True)
         if choice is None:
             return None
         do_clone = choice
@@ -228,7 +228,7 @@ def _execute(settings: dict) -> int:
     if voices is None:
         return 1
 
-    # Sync converter/config.py port + default voice.
+    # Sync app/converter/config.py port + default voice.
     port = settings["port"]
     new_url = common.url_with_port(config.FASTER_API_URL, port)
     if new_url != config.FASTER_API_URL:
@@ -236,14 +236,14 @@ def _execute(settings: dict) -> int:
             print(f"[OK] Updated FASTER_API_URL to {new_url}")
         else:
             print("[WARNING] Could not update FASTER_API_URL; edit "
-                  "converter/config.py by hand")
+                  "app/converter/config.py by hand")
     default_voice = next(iter(voices))
     if default_voice != config.FASTER_VOICE:
         if common.update_config_value("FASTER_VOICE", default_voice):
             print(f"[OK] Updated FASTER_VOICE to {default_voice}")
         else:
             print("[WARNING] Could not update FASTER_VOICE; edit "
-                  "converter/config.py by hand")
+                  "app/converter/config.py by hand")
 
     _print_launch_hint(settings["output_path"], port)
     return 0
@@ -312,14 +312,14 @@ def _collect_from_flags(args: argparse.Namespace,
 def build_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(
         description="Set up the faster-qwen3-tts backend: pip install, clone, "
-                    "build voices.json, and sync converter/config.py.")
+                    "build voices.json, and sync app/converter/config.py.")
     parser.add_argument("input_dir", type=normalize_dir_arg, nargs="?",
                         default=None, metavar="WAV_DIR",
                         help="Directory with .wav reference files "
                              f"(default: {VOICES_DIR}; browsed for in the TUI)")
     parser.add_argument("--output", type=Path, default=None,
                         help="Output path for voices.json (default: "
-                             "./faster-qwen3-tts/voices.json, or "
+                             "./app/faster-qwen3-tts/voices.json, or "
                              "WAV_DIR/voices.json when not cloned)")
     parser.add_argument("--language", type=str, default=None,
                         help="Language for all voices (default: English; "
@@ -332,7 +332,7 @@ def build_parser() -> argparse.ArgumentParser:
                         help="Overwrite an existing voices.json without "
                              "prompting")
     parser.add_argument("--port", type=int, default=None,
-                        help="Server port to record in converter/config.py "
+                        help="Server port to record in app/converter/config.py "
                              "(default: the port in FASTER_API_URL)")
     parser.add_argument("--skip-install", action="store_true",
                         help="Do not pip install faster-qwen3-tts[demo]")
@@ -352,7 +352,7 @@ def detect() -> BackendStatus:
     details.append("pip: installed" if installed else
                    "not installed — run setup to pip install")
     details.append(f"checkout: {_checkout()}" if cloned else
-                   f"not cloned — run setup to clone ./{FASTER_DIR_NAME}")
+                   f"not cloned — run setup to clone ./app/{FASTER_DIR_NAME}")
     details.append(f"voices: {voices_json}" if voices_json.exists() else
                    "no voices.json — run setup to create one")
     launch = ""
diff --git a/backends/qwen.py b/app/backends/qwen.py
index 52f7a3f..21280a0 100644
--- a/backends/qwen.py
+++ b/app/backends/qwen.py
@@ -5,11 +5,11 @@ qwen-tts is a pip package providing the ``qwen-tts-demo`` server, which
 hosts the Qwen3-TTS CustomVoice (built-in speakers) and Base (voice
 cloning) models on separate ports. This module sets it up end-to-end as a
 TUI: pip-install the package, configure the two ports and the built-in
-speaker in ``converter/config.py``, and print the launch commands. It is
+speaker in ``app/converter/config.py``, and print the launch commands. It is
 driven by ``audiobook.py``'s hub but can also be run directly with flags.
 
 Usage:
-    python -m backends.qwen [--port-custom PORT] [--port-clone PORT]
+    python app/backends/qwen.py [--port-custom PORT] [--port-clone PORT]
         [--speaker NAME] [--skip-install]
 """
 
@@ -37,7 +37,7 @@ QWEN_BASE_MODEL = "Qwen/Qwen3-TTS-12Hz-1.7B-Base"
 DEFAULT_CUSTOM_PORT = 7860
 DEFAULT_CLONE_PORT = 7861
 
-# Built-in CustomVoice speakers (see converter/config.py SPEAKER).
+# Built-in CustomVoice speakers (see app/converter/config.py SPEAKER).
 QWEN_SPEAKERS = ("Vivian", "Serena", "Uncle_Fu", "Dylan", "Eric", "Ryan",
                  "Aiden", "Ono_Anna", "Sohee")
 
@@ -128,20 +128,20 @@ def _execute(settings: dict) -> int:
             print(f"[OK] Updated QWEN_API_URL to {custom_url}")
         else:
             print("[WARNING] Could not update QWEN_API_URL; edit "
-                  "converter/config.py by hand")
+                  "app/converter/config.py by hand")
     clone_url = common.url_with_port(config.CLONE_API_URL, settings["clone_port"])
     if clone_url != config.CLONE_API_URL:
         if common.update_config_value("CLONE_API_URL", clone_url):
             print(f"[OK] Updated CLONE_API_URL to {clone_url}")
         else:
             print("[WARNING] Could not update CLONE_API_URL; edit "
-                  "converter/config.py by hand")
+                  "app/converter/config.py by hand")
     if settings["speaker"] != config.SPEAKER:
         if common.update_config_value("SPEAKER", settings["speaker"]):
             print(f"[OK] Updated SPEAKER to {settings['speaker']}")
         else:
             print("[WARNING] Could not update SPEAKER; edit "
-                  "converter/config.py by hand")
+                  "app/converter/config.py by hand")
 
     _print_launch_hint(settings["custom_port"], settings["clone_port"])
     return 0
diff --git a/backends/servers.py b/app/backends/servers.py
index 12846f0..a5a6829 100644
--- a/backends/servers.py
+++ b/app/backends/servers.py
@@ -4,13 +4,13 @@ Each backend's ``detect()`` returns a list of ``ServerSpec`` — the exact argv
 (absolute binaries in the managed venv, no shell activation needed) and the
 URL to probe for readiness. This module turns those specs into running
 processes: ``start`` spawns the server, streams its output to
-``logs/<name>-server.log``, records its pid, and polls the URL until it
+``app/logs/<name>-server.log``, records its pid, and polls the URL until it
 accepts connections (model loads are slow, so the timeout is generous);
 ``stop`` terminates the process group the hub started.
 
 Everything here runs in the plain console tail after the curses TUI returns
 (matching the wizards' build/pip streaming), so progress and log tails appear
-normally. Pid/log files live under ``logs/`` which is already gitignored.
+normally. Pid/log files live under ``app/logs/`` which is already gitignored.
 """
 
 import os
@@ -22,9 +22,9 @@ from pathlib import Path
 from typing import List
 
 from backends import common
-from backends.common import TTS_ROOT
+from backends.common import APP_DIR
 
-LOG_DIR = TTS_ROOT / "logs"
+LOG_DIR = APP_DIR / "logs"
 
 # How long to wait for a server to accept connections on its URL. First-time
 # model loads (especially qwen-tts / faster-qwen3-tts pulling weights into
diff --git a/converter/__init__.py b/app/converter/__init__.py
index 86a827f..86a827f 100644
--- a/converter/__init__.py
+++ b/app/converter/__init__.py
diff --git a/converter/audio.py b/app/converter/audio.py
index d1dc28b..81431cb 100644
--- a/converter/audio.py
+++ b/app/converter/audio.py
@@ -13,7 +13,7 @@ from . import config
 
 logger = logging.getLogger(__name__)
 
-CHUNKS_FOLDER = Path(__file__).resolve().parent.parent / "chunks"
+CHUNKS_FOLDER = Path(__file__).resolve().parent.parent.parent / "app" / "chunks"
 
 
 def atempo_filters(speed: float) -> str:
diff --git a/converter/chunking.py b/app/converter/chunking.py
index 800b76a..9ef6a5d 100644
--- a/converter/chunking.py
+++ b/app/converter/chunking.py
@@ -16,7 +16,7 @@ def split_into_chunks(text: str, max_words: Optional[int] = None) -> List[str]:
     There is no ceiling beyond that setting, but note that the TTS
     servers silently truncate audio when a single generation runs too
     long without reporting an error, so very large values are at your
-    own risk (see CHUNK_SIZE in converter/config.py).
+    own risk (see CHUNK_SIZE in app/converter/config.py).
 
     Splits on sentence boundaries. Sentences longer than the limit are
     split further at clause punctuation (which is kept attached for TTS
diff --git a/converter/config.py b/app/converter/config.py
index 6a98136..6a98136 100644
--- a/converter/config.py
+++ b/app/converter/config.py
diff --git a/converter/converter.py b/app/converter/converter.py
index b24ac09..cef4808 100644
--- a/converter/converter.py
+++ b/app/converter/converter.py
@@ -33,14 +33,16 @@ from .tts import (
 logger = logging.getLogger(__name__)
 
 # Folders, resolved from the project root so the converter runs from any
-# working directory.
-BASE_DIR = Path(__file__).resolve().parent.parent
+# working directory. User-facing dirs (input/, output/) stay at the root;
+# scratch/log dirs live under the app/ container.
+BASE_DIR = Path(__file__).resolve().parent.parent.parent
+APP_DIR = BASE_DIR / "app"
 
 BOOKS_FOLDER = BASE_DIR / "input"
 AUDIOBOOKS_FOLDER = BASE_DIR / "output"
-CHUNKS_FOLDER = BASE_DIR / "chunks"  # Per-chunk scratch audio, cleaned per book
-LOGS_FOLDER = BASE_DIR / "logs"
-DEBUG_FOLDER = BASE_DIR / "debug"  # --debug dumps, kept across runs
+CHUNKS_FOLDER = APP_DIR / "chunks"  # Per-chunk scratch audio, cleaned per book
+LOGS_FOLDER = APP_DIR / "logs"
+DEBUG_FOLDER = APP_DIR / "debug"  # --debug dumps, kept across runs
 
 # Output containers and supported input formats.
 AUDIO_FORMATS = ("mp3", "m4b", "ogg", "flac")
diff --git a/converter/cover.py b/app/converter/cover.py
index b2d3cb5..b2d3cb5 100644
--- a/converter/cover.py
+++ b/app/converter/cover.py
diff --git a/converter/extractors.py b/app/converter/extractors.py
index a564333..a564333 100644
--- a/converter/extractors.py
+++ b/app/converter/extractors.py
diff --git a/converter/tts.py b/app/converter/tts.py
index f5d54e2..6ccef56 100644
--- a/converter/tts.py
+++ b/app/converter/tts.py
@@ -180,7 +180,7 @@ MODEL_SIZE = "1.7B"
 CUSTOM_VOICE_MODEL_ID = "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice"
 SAMPLE_RATE = 24000
 
-CHUNKS_FOLDER = Path(__file__).resolve().parent.parent / "chunks"
+CHUNKS_FOLDER = Path(__file__).resolve().parent.parent.parent / "app" / "chunks"
 
 
 def _resolve_request_seed() -> int:
@@ -990,7 +990,7 @@ class AudioCppTTSClient(_BaseTTSClient):
         """Resolve an empty model id when the server hosts exactly one entry.
 
         Multi-model servers generated with several lazily-loaded entries can
-        be used without editing converter/config.py: leave AUDIOCPP_MODEL_ID
+        be used without editing app/converter/config.py: leave AUDIOCPP_MODEL_ID
         (and ``--model``) unset, and the single hosted entry is chosen
         automatically. With more than one entry an explicit choice is required
         (via ``--model`` or AUDIOCPP_MODEL_ID), since guessing would risk
@@ -1025,7 +1025,7 @@ class AudioCppTTSClient(_BaseTTSClient):
                 f"The audio.cpp server at {self.api_url} hosts {len(model_ids)} "
                 f"model entries ({configured}); audiobook.py needs to know which "
                 "one to use. Pass --model <id> when converting, or set "
-                "AUDIOCPP_MODEL_ID in converter/config.py to one of them "
+                "AUDIOCPP_MODEL_ID in app/converter/config.py to one of them "
                 "(see README)."
             )
         if self.preset_mode:
@@ -1035,14 +1035,14 @@ class AudioCppTTSClient(_BaseTTSClient):
                 f"'{config.AUDIOCPP_CLONE_MODEL_ID}' (configured: {configured}). "
                 "Add a TTS model entry for the family you want to the server "
                 "config and match AUDIOCPP_MODEL_ID / AUDIOCPP_CLONE_MODEL_ID "
-                "in converter/config.py to its id, or select it per run with "
+                "in app/converter/config.py to its id, or select it per run with "
                 "--model (see README)."
             )
         raise RuntimeError(
             f"The audio.cpp server at {self.api_url} has no model id "
             f"'{self.model_id}' (configured: {configured}). Speaker mode needs "
             "the Qwen3-TTS CustomVoice model: add a qwen3_tts model entry to "
-            "the server config and match AUDIOCPP_MODEL_ID in converter/config.py to its "
+            "the server config and match AUDIOCPP_MODEL_ID in app/converter/config.py to its "
             "id (or pass --model), or rerun with --voice to use a voice preset "
             "on any TTS model (see README)."
         )
@@ -1088,7 +1088,7 @@ class AudioCppTTSClient(_BaseTTSClient):
                 "AUDIOCPP_CLONE_MODEL_ID %r hosts family %r, but "
                 "AUDIOCPP_MODEL_ID %r hosts %r; preset requests stay on "
                 "'%s'. Point both ids at the same model entry in "
-                "converter/config.py (single-model servers use the same id "
+                "app/converter/config.py (single-model servers use the same id "
                 "for both)",
                 clone_model_id, clone_family, self.model_id, primary_family,
                 self.model_id)
diff --git a/docs/backend-audiocpp.md b/app/docs/backend-audiocpp.md
index ee511bb..271c9b6 100644
--- a/docs/backend-audiocpp.md
+++ b/app/docs/backend-audiocpp.md
@@ -2,7 +2,7 @@
 
 `--backend audiocpp` talks to `audiocpp_server` from [audio.cpp](https://github.com/0xShug0/audio.cpp), which hosts numerous TTS model families.
 
-The easiest way is the TUI: run `python audiobook.py`, choose **Set up a backend… → audio.cpp**, and it clones `audio.cpp` into `./audio.cpp` (or reuses an existing checkout), builds `audiocpp_server`, lets you pick model families/packages from an expandable checkbox tree (reading the checkout's `model_specs/`), transcribes `.wav` voices with `whisper`, writes `server.json` into the checkout, syncs `converter/config.py`, and prints the launch command (the hub can also start the server for you via the **Server** menu or automatically when converting). Run it directly with `python -m backends.audiocpp` (flags like `--wavs`, `--families`, `--build-backend`, `--clone` skip the corresponding screens for scripting). The TUI runs in the managed `envs/tts` venv, which includes `whisper` via `requirements.txt`; for a manual setup, make sure `whisper` (or `faster_whisper`) is installed in the environment you run the wizard from. The Qwen3-TTS model tree also offers hosting the VoiceDesign package as a `vdes` entry.
+The easiest way is the TUI: run `python audiobook.py`, choose **Set up a backend… → audio.cpp**, and it clones `audio.cpp` into `app/audio.cpp` (or reuses an existing checkout), builds `audiocpp_server`, lets you pick model families/packages from an expandable checkbox tree (reading the checkout's `model_specs/`), transcribes `.wav` voices with `whisper`, writes `server.json` into the checkout, syncs `app/converter/config.py`, and prints the launch command (the hub can also start the server for you via the **Server** menu or automatically when converting). Run it directly with `python app/backends/audiocpp.py` (flags like `--wavs`, `--families`, `--build-backend`, `--clone` skip the corresponding screens for scripting). The TUI runs in the managed `app/envs/tts` venv, which includes `whisper` via `requirements.txt`; for a manual setup, make sure `whisper` (or `faster_whisper`) is installed in the environment you run the wizard from. The Qwen3-TTS model tree also offers hosting the VoiceDesign package as a `vdes` entry.
 
 If you prefer to install the backend yourself (in your own environment, not the managed venv), the manual steps are below. Either way the hub detects a running server by its port, so a manually-installed backend works once its server is up.
 
diff --git a/docs/backend-faster.md b/app/docs/backend-faster.md
index c407a70..b08e057 100644
--- a/docs/backend-faster.md
+++ b/app/docs/backend-faster.md
@@ -2,7 +2,7 @@
 
 `--backend faster` talks to the OpenAI-compatible server from [faster-qwen3-tts](https://github.com/andimarafioti/faster-qwen3-tts), which uses CUDA graph capture for roughly 5-10x faster inference with the same models. **It requires an NVIDIA GPU**.
 
-The easiest way is to run `python audiobook.py` → **Set up a backend… → faster-qwen3-tts** (or `python -m backends.faster path/to/clone/wavs`): the TUI pip-installs `faster-qwen3-tts[demo]` into its managed venv (`envs/tts`), clones the repo, transcribes the `.wav` files with `whisper`, and writes `voices.json` for you. You can also start the server from the hub's **Server** menu, or let a conversion start it automatically.
+The easiest way is to run `python audiobook.py` → **Set up a backend… → faster-qwen3-tts** (or `python app/backends/faster.py path/to/clone/wavs`): the TUI pip-installs `faster-qwen3-tts[demo]` into its managed venv (`app/envs/tts`), clones the repo, transcribes the `.wav` files with `whisper`, and writes `voices.json` for you. You can also start the server from the hub's **Server** menu, or let a conversion start it automatically.
 
 If you prefer to install the backend yourself (in your own environment, not the managed venv), the manual steps are below. Either way the hub detects a running server by its port, so a manually-installed backend works once its server is up.
 
diff --git a/docs/backend-qwen.md b/app/docs/backend-qwen.md
index 028d6f3..074004c 100644
--- a/docs/backend-qwen.md
+++ b/app/docs/backend-qwen.md
@@ -1,6 +1,6 @@
 # Backend Option 2: Qwen3-TTS
 
-The easiest way is to run `python audiobook.py` → **Set up a backend… → qwen-tts** (or `python -m backends.qwen`): the TUI pip-installs `qwen-tts` into its managed venv (`envs/tts`), configures the two ports and the built-in speaker in `converter/config.py`, and prints the launch commands. You can also start the server from the hub's **Server** menu, or let a conversion start it automatically.
+The easiest way is to run `python audiobook.py` → **Set up a backend… → qwen-tts** (or `python app/backends/qwen.py`): the TUI pip-installs `qwen-tts` into its managed venv (`app/envs/tts`), configures the two ports and the built-in speaker in `app/converter/config.py`, and prints the launch commands. You can also start the server from the hub's **Server** menu, or let a conversion start it automatically.
 
 If you prefer to install the backend yourself (in your own environment, not the managed venv), the manual steps are below. Either way the hub detects a running server by its port, so a manually-installed backend works once its server is up.
 
@@ -40,7 +40,7 @@ qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --ip 127.0.0.1 --port 7860 [-
 python audiobook.py --backend qwen
 ```
 
-Change the voice settings in `converter/config.py`.
+Change the voice settings in `app/converter/config.py`.
 
 ## Optional: FlashAttention for qwen-tts-demo server
 
diff --git a/tests/__init__.py b/app/tests/__init__.py
index e69de29..e69de29 100644
--- a/tests/__init__.py
+++ b/app/tests/__init__.py
diff --git a/tests/cover_test.png b/app/tests/cover_test.png
index 0c252db..0c252db 100644
--- a/tests/cover_test.png
+++ b/app/tests/cover_test.png
diff --git a/tests/gen_test_cover.py b/app/tests/gen_test_cover.py
index 292469a..292469a 100644
--- a/tests/gen_test_cover.py
+++ b/app/tests/gen_test_cover.py
diff --git a/tests/test_audio.py b/app/tests/test_audio.py
index ef5e92a..ef5e92a 100644
--- a/tests/test_audio.py
+++ b/app/tests/test_audio.py
diff --git a/tests/test_backends.py b/app/tests/test_backends.py
index c0e8d4a..c0e8d4a 100644
--- a/tests/test_backends.py
+++ b/app/tests/test_backends.py
diff --git a/tests/test_backends_audiocpp.py b/app/tests/test_backends_audiocpp.py
index 9882ce1..9882ce1 100644
--- a/tests/test_backends_audiocpp.py
+++ b/app/tests/test_backends_audiocpp.py
diff --git a/tests/test_backends_envs.py b/app/tests/test_backends_envs.py
index cf4ecc6..cf4ecc6 100644
--- a/tests/test_backends_envs.py
+++ b/app/tests/test_backends_envs.py
diff --git a/tests/test_backends_faster.py b/app/tests/test_backends_faster.py
index 641f6ee..641f6ee 100644
--- a/tests/test_backends_faster.py
+++ b/app/tests/test_backends_faster.py
diff --git a/tests/test_backends_servers.py b/app/tests/test_backends_servers.py
index 02b65e6..02b65e6 100644
--- a/tests/test_backends_servers.py
+++ b/app/tests/test_backends_servers.py
diff --git a/tests/test_chunking.py b/app/tests/test_chunking.py
index 2904e40..2904e40 100644
--- a/tests/test_chunking.py
+++ b/app/tests/test_chunking.py
diff --git a/tests/test_cleaning.py b/app/tests/test_cleaning.py
index 41f4ed7..41f4ed7 100644
--- a/tests/test_cleaning.py
+++ b/app/tests/test_cleaning.py
diff --git a/tests/test_converter.py b/app/tests/test_converter.py
index 2fe0f5d..2fe0f5d 100644
--- a/tests/test_converter.py
+++ b/app/tests/test_converter.py
diff --git a/tests/test_cover.py b/app/tests/test_cover.py
index f19db5a..f19db5a 100644
--- a/tests/test_cover.py
+++ b/app/tests/test_cover.py
diff --git a/tests/test_extractors.py b/app/tests/test_extractors.py
index ae1794c..ae1794c 100644
--- a/tests/test_extractors.py
+++ b/app/tests/test_extractors.py
diff --git a/tests/test_hub.py b/app/tests/test_hub.py
index b3a17e8..b3a17e8 100644
--- a/tests/test_hub.py
+++ b/app/tests/test_hub.py
diff --git a/tests/test_tts.py b/app/tests/test_tts.py
index a2df07f..a2df07f 100644
--- a/tests/test_tts.py
+++ b/app/tests/test_tts.py
diff --git a/tests/test_tui.py b/app/tests/test_tui.py
index c121d55..c121d55 100644
--- a/tests/test_tui.py
+++ b/app/tests/test_tui.py
diff --git a/ui/__init__.py b/app/ui/__init__.py
index 4986a10..4986a10 100644
--- a/ui/__init__.py
+++ b/app/ui/__init__.py
diff --git a/ui/hub.py b/app/ui/hub.py
index b2a2b08..6a94096 100644
--- a/ui/hub.py
+++ b/app/ui/hub.py
@@ -200,7 +200,7 @@ def _convert_menu(stdscr, statuses) -> Optional[tuple]:
 
 
 def _convert_audiocpp(stdscr, statuses) -> Optional[tuple]:
-    """Collect audio.cpp run settings by reading ./audio.cpp/server.json."""
+    """Collect audio.cpp run settings by reading app/audio.cpp/server.json."""
     checkout = audiocpp_backend.find_local_checkout()
     server_json = checkout / "server.json" if checkout else None
     if not server_json or not server_json.exists():
@@ -379,7 +379,7 @@ def _common_options(stdscr) -> Optional[dict]:
 
 
 # ---------------------------------------------------------------------------
-# Settings menu (global output options -> converter/config.py)
+# Settings menu (global output options -> app/converter/config.py)
 # ---------------------------------------------------------------------------
 
 def _settings_menu(stdscr) -> None:
@@ -434,7 +434,7 @@ def _validate_chunk_size(value: str) -> Optional[str]:
 
 
 def _apply_settings(values: dict) -> None:
-    """Write VALUES to converter/config.py and reload them in-memory."""
+    """Write VALUES to app/converter/config.py and reload them in-memory."""
     chunk_size = int(values["chunk_size"].strip())
     if chunk_size < 1:
         raise ValueError("Chunk size must be at least 1")
@@ -455,7 +455,7 @@ def _apply_settings(values: dict) -> None:
 
 
 def _write_config(updates: dict) -> None:
-    """Rewrite the ``NAME = value`` lines for UPDATES in converter/config.py.
+    """Rewrite the ``NAME = value`` lines for UPDATES in app/converter/config.py.
 
     Only the value of each named assignment changes: the indentation, the
     quotes (double, matching the file's style) and any trailing comment on
diff --git a/ui/tui.py b/app/ui/tui.py
index 9047f36..9047f36 100644
--- a/ui/tui.py
+++ b/app/ui/tui.py
diff --git a/audiobook.py b/audiobook.py
index 792b274..6c075ea 100755
--- a/audiobook.py
+++ b/audiobook.py
@@ -5,12 +5,13 @@ Converts TXT, PDF and EPUB files into audiobooks using a local TTS server.
 
 Run with no arguments in a terminal for the full TUI (set up backends,
 process the input directory); pass flags to script a conversion directly.
-Edit converter/config.py to change voice and processing settings.
+Edit app/converter/config.py to change voice and processing settings.
 """
 
 import argparse
 import sys
 import traceback
+from pathlib import Path
 
 # Fix Windows console encoding for unicode output
 if sys.platform == "win32":
@@ -20,6 +21,12 @@ if sys.platform == "win32":
     except AttributeError:
         pass
 
+# Everything non-user-facing (source packages, generated dirs, venv, backend
+# checkouts) lives under ./app so the checkout root stays clean. Put it on
+# sys.path before importing the packages below.
+APP_DIR = Path(__file__).resolve().parent / "app"
+sys.path.insert(0, str(APP_DIR))
+
 # The managed-environment bootstrap (backends.envs) is stdlib-only and is
 # imported here so main() can launch it before any third-party dependency is
 # touched. It must NOT run at import time (importing this module must stay
@@ -174,7 +181,7 @@ Examples:
               "With --backend audiocpp the language is adapted to the model "
               "family: sent as a code (e.g. 'en') for families that take one, or "
               "omitted when the model detects the language itself. Defaults to "
-              "the LANGUAGE setting in converter/config.py (English).")
+              "the LANGUAGE setting in app/converter/config.py (English).")
     )
     parser.add_argument(
         "--speed", type=float, default=1.0,
@@ -198,7 +205,7 @@ Examples:
               "audio.cpp audiocpp_server (audiocpp) hosting any of its TTS "
               "model families — Qwen3-TTS, Higgs Audio, VoxCPM2, IndexTTS2, "
               "and more. Defaults to the BACKEND setting in "
-              "converter/config.py (audiocpp).")
+              "app/converter/config.py (audiocpp).")
     )
     parser.add_argument(
         "--voice", type=str, default=None, metavar="NAME",
@@ -207,7 +214,7 @@ Examples:
               "with --ref-audio). audiocpp: a voice_preset or voice_dir entry "
               "(cloning); required for audio.cpp families without built-in "
               "speakers (everything except Qwen3-TTS CustomVoice). Not used by "
-              "the qwen backend (use converter/config.py SPEAKER or --clone "
+              "the qwen backend (use app/converter/config.py SPEAKER or --clone "
               "there).")
     )
     parser.add_argument(
@@ -219,7 +226,7 @@ Examples:
     parser.add_argument(
         "--chunk", action="store_true",
         help=("Force client-side chunking into CHUNK_SIZE-word requests (see "
-              "converter/config.py). Only matters for --backend audiocpp, which "
+              "app/converter/config.py). Only matters for --backend audiocpp, which "
               "otherwise sends each chapter as one request and lets the server "
               "chunk long text itself; the qwen and faster backends always "
               "chunk.")
@@ -228,7 +235,7 @@ Examples:
         "--model", type=str, default=None, metavar="ID",
         help=("audio.cpp server model entry id to use for this run "
               "(--backend audiocpp only). Overrides AUDIOCPP_MODEL_ID in "
-              "converter/config.py, which is useful for a server hosting "
+              "app/converter/config.py, which is useful for a server hosting "
               "several lazily-loaded models: generate one server.json with "
               "backends.audiocpp, then pick the model per run with --model. "
               "Leave unset to use the config id, or to auto-select when the "
@@ -243,7 +250,7 @@ Examples:
               "'A warm adult female narrator with a British accent'. On "
               "other families it acts as a style/delivery instruction when "
               "the model supports one and is ignored otherwise. Defaults to "
-              "AUDIOCPP_INSTRUCTIONS in converter/config.py (empty).")
+              "AUDIOCPP_INSTRUCTIONS in app/converter/config.py (empty).")
     )
     parser.add_argument(
         "--option", action="append", type=str, default=None, metavar="KEY=VALUE",
@@ -306,7 +313,7 @@ Examples:
         if args.voice is not None:
             parser.error("--voice requires --backend faster or audiocpp; the "
                          "qwen backend uses built-in speakers "
-                         "(converter/config.py SPEAKER) or --clone")
+                         "(app/converter/config.py SPEAKER) or --clone")
         if args.language is not None:
             try:
                 args.language = normalize_language(args.language)