diff --git a/backend/app.py b/backend/app.py
index f8bbbcd1..69cee59b 100644
--- a/backend/app.py
+++ b/backend/app.py
@@ -15,9 +15,10 @@ from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 from urllib.parse import quote
 
-from . import __version__, database, tts, transcribe, config
+from . import __version__, database
+from .services import tts, transcribe
 from .database import get_db
-from .platform_detect import get_backend_type
+from .utils.platform_detect import get_backend_type
 from .utils.progress import get_progress_manager
 from .services.task_queue import create_background_task, init_queue
 from .routes import register_routers
diff --git a/backend/backends/__init__.py b/backend/backends/__init__.py
index 3ebbb985..cc35eabe 100644
--- a/backend/backends/__init__.py
+++ b/backend/backends/__init__.py
@@ -11,7 +11,7 @@ from typing import Protocol, Optional, Tuple, List
 from typing_extensions import runtime_checkable
 import numpy as np
 
-from ..platform_detect import get_backend_type
+from ..utils.platform_detect import get_backend_type
 
 LANGUAGE_CODE_TO_NAME = {
     "zh": "chinese",
@@ -375,7 +375,7 @@ async def ensure_model_cached_or_raise(engine: str, model_size: str = "default")
 def unload_model_by_config(config: ModelConfig) -> bool:
     """Unload a model given its config. Returns True if it was loaded, False otherwise."""
     from . import get_tts_backend_for_engine
-    from .. import tts, transcribe
+    from ..services import tts, transcribe
 
     if config.engine == "whisper":
         whisper_model = transcribe.get_whisper_model()
@@ -403,7 +403,7 @@ def unload_model_by_config(config: ModelConfig) -> bool:
 def check_model_loaded(config: ModelConfig) -> bool:
     """Check if a model is currently loaded."""
     from . import get_tts_backend_for_engine
-    from .. import tts, transcribe
+    from ..services import tts, transcribe
 
     try:
         if config.engine == "whisper":
@@ -424,7 +424,7 @@ def check_model_loaded(config: ModelConfig) -> bool:
 def get_model_load_func(config: ModelConfig):
     """Return a callable that loads/downloads the model."""
     from . import get_tts_backend_for_engine
-    from .. import tts, transcribe
+    from ..services import tts, transcribe
 
     if config.engine == "whisper":
         return lambda: transcribe.get_whisper_model().load_model(config.model_size)
diff --git a/backend/build_binary.py b/backend/build_binary.py
index 43de991b..6f8b1736 100644
--- a/backend/build_binary.py
+++ b/backend/build_binary.py
@@ -55,11 +55,11 @@ def build_server(cuda=False):
         '--hidden-import', 'backend.config',
         '--hidden-import', 'backend.database',
         '--hidden-import', 'backend.models',
-        '--hidden-import', 'backend.profiles',
-        '--hidden-import', 'backend.history',
-        '--hidden-import', 'backend.tts',
-        '--hidden-import', 'backend.transcribe',
-        '--hidden-import', 'backend.platform_detect',
+        '--hidden-import', 'backend.services.profiles',
+        '--hidden-import', 'backend.services.history',
+        '--hidden-import', 'backend.services.tts',
+        '--hidden-import', 'backend.services.transcribe',
+        '--hidden-import', 'backend.utils.platform_detect',
         '--hidden-import', 'backend.backends',
         '--hidden-import', 'backend.backends.pytorch_backend',
         '--hidden-import', 'backend.utils.audio',
@@ -68,9 +68,9 @@ def build_server(cuda=False):
         '--hidden-import', 'backend.utils.hf_progress',
         '--hidden-import', 'backend.utils.validation',
         '--hidden-import', 'backend.services.cuda',
-        '--hidden-import', 'backend.effects',
+        '--hidden-import', 'backend.services.effects',
         '--hidden-import', 'backend.utils.effects',
-        '--hidden-import', 'backend.versions',
+        '--hidden-import', 'backend.services.versions',
         '--hidden-import', 'pedalboard',
         '--hidden-import', 'chatterbox',
         '--hidden-import', 'chatterbox.tts_turbo',
diff --git a/backend/routes/audio.py b/backend/routes/audio.py
index 12e774ef..682d7aae 100644
--- a/backend/routes/audio.py
+++ b/backend/routes/audio.py
@@ -6,7 +6,8 @@ from fastapi import APIRouter, Depends, HTTPException
 from fastapi.responses import FileResponse
 from sqlalchemy.orm import Session
 
-from .. import history, models
+from .. import models
+from ..services import history
 from ..database import get_db
 
 router = APIRouter()
@@ -15,7 +16,7 @@ router = APIRouter()
 @router.get("/audio/version/{version_id}")
 async def get_version_audio(version_id: str, db: Session = Depends(get_db)):
     """Serve audio for a specific version."""
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
 
     version = versions_mod.get_version(version_id, db)
     if not version:
diff --git a/backend/routes/channels.py b/backend/routes/channels.py
index 56391624..c13162fb 100644
--- a/backend/routes/channels.py
+++ b/backend/routes/channels.py
@@ -3,7 +3,8 @@
 from fastapi import APIRouter, Depends, HTTPException
 from sqlalchemy.orm import Session
 
-from .. import channels, models
+from .. import models
+from ..services import channels
 from ..database import get_db
 
 router = APIRouter()
diff --git a/backend/routes/effects.py b/backend/routes/effects.py
index 8e1f6272..8139176d 100644
--- a/backend/routes/effects.py
+++ b/backend/routes/effects.py
@@ -9,7 +9,8 @@ from fastapi import APIRouter, Depends, HTTPException
 from fastapi.responses import StreamingResponse
 from sqlalchemy.orm import Session
 
-from .. import config, history, models
+from .. import config, models
+from ..services import history
 from ..database import Generation as DBGeneration, get_db
 
 router = APIRouter()
@@ -28,7 +29,7 @@ async def preview_effects(
     if (gen.status or "completed") != "completed":
         raise HTTPException(status_code=400, detail="Generation is not completed")
 
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
     from ..utils.effects import apply_effects, validate_effects_chain
     from ..utils.audio import load_audio
 
@@ -73,7 +74,7 @@ async def get_available_effects():
 @router.get("/effects/presets", response_model=list[models.EffectPresetResponse])
 async def list_effect_presets(db: Session = Depends(get_db)):
     """List all effect presets (built-in + user-created)."""
-    from .. import effects as effects_mod
+    from ..services import effects as effects_mod
 
     return effects_mod.list_presets(db)
 
@@ -81,7 +82,7 @@ async def list_effect_presets(db: Session = Depends(get_db)):
 @router.get("/effects/presets/{preset_id}", response_model=models.EffectPresetResponse)
 async def get_effect_preset(preset_id: str, db: Session = Depends(get_db)):
     """Get a specific effect preset."""
-    from .. import effects as effects_mod
+    from ..services import effects as effects_mod
 
     preset = effects_mod.get_preset(preset_id, db)
     if not preset:
@@ -95,7 +96,7 @@ async def create_effect_preset(
     db: Session = Depends(get_db),
 ):
     """Create a new effect preset."""
-    from .. import effects as effects_mod
+    from ..services import effects as effects_mod
 
     try:
         return effects_mod.create_preset(data, db)
@@ -110,7 +111,7 @@ async def update_effect_preset(
     db: Session = Depends(get_db),
 ):
     """Update an effect preset."""
-    from .. import effects as effects_mod
+    from ..services import effects as effects_mod
 
     try:
         result = effects_mod.update_preset(preset_id, data, db)
@@ -124,7 +125,7 @@ async def update_effect_preset(
 @router.delete("/effects/presets/{preset_id}")
 async def delete_effect_preset(preset_id: str, db: Session = Depends(get_db)):
     """Delete a user effect preset."""
-    from .. import effects as effects_mod
+    from ..services import effects as effects_mod
 
     try:
         if not effects_mod.delete_preset(preset_id, db):
@@ -147,7 +148,7 @@ async def list_generation_versions(
     if not gen:
         raise HTTPException(status_code=404, detail="Generation not found")
 
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
 
     return versions_mod.list_versions(generation_id, db)
 
@@ -168,7 +169,7 @@ async def apply_effects_to_generation(
     if (gen.status or "completed") != "completed":
         raise HTTPException(status_code=400, detail="Generation is not completed")
 
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
     from ..utils.effects import apply_effects, validate_effects_chain
     from ..utils.audio import load_audio, save_audio
 
@@ -227,7 +228,7 @@ async def set_default_version(
     db: Session = Depends(get_db),
 ):
     """Set a specific version as the default for a generation."""
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
 
     version = versions_mod.get_version(version_id, db)
     if not version or version.generation_id != generation_id:
@@ -246,7 +247,7 @@ async def delete_generation_version(
     db: Session = Depends(get_db),
 ):
     """Delete a version. Cannot delete the last remaining version."""
-    from .. import versions as versions_mod
+    from ..services import versions as versions_mod
 
     version = versions_mod.get_version(version_id, db)
     if not version or version.generation_id != generation_id:
diff --git a/backend/routes/generations.py b/backend/routes/generations.py
index 32051796..d9f4ef29 100644
--- a/backend/routes/generations.py
+++ b/backend/routes/generations.py
@@ -7,7 +7,8 @@ from fastapi import APIRouter, Depends, HTTPException
 from fastapi.responses import StreamingResponse
 from sqlalchemy.orm import Session
 
-from .. import history, models, profiles, tts
+from .. import models
+from ..services import history, profiles, tts
 from ..database import Generation as DBGeneration, VoiceProfile as DBVoiceProfile, get_db
 from ..services.generation import run_generation
 from ..services.task_queue import enqueue_generation
diff --git a/backend/routes/health.py b/backend/routes/health.py
index 181aaebf..e48d5689 100644
--- a/backend/routes/health.py
+++ b/backend/routes/health.py
@@ -8,9 +8,10 @@ import torch
 from fastapi import APIRouter, Depends
 from sqlalchemy.orm import Session
 
-from .. import config, models, tts
+from .. import config, models
+from ..services import tts
 from ..database import get_db
-from ..platform_detect import get_backend_type
+from ..utils.platform_detect import get_backend_type
 
 router = APIRouter()
 
diff --git a/backend/routes/history.py b/backend/routes/history.py
index e4c2d47c..5ae438b2 100644
--- a/backend/routes/history.py
+++ b/backend/routes/history.py
@@ -7,7 +7,8 @@ from fastapi import APIRouter, Depends, File, HTTPException, UploadFile
 from fastapi.responses import FileResponse, StreamingResponse
 from sqlalchemy.orm import Session
 
-from .. import export_import, history, models
+from .. import models
+from ..services import export_import, history
 from ..app import safe_content_disposition
 from ..database import Generation as DBGeneration, VoiceProfile as DBVoiceProfile, get_db
 
diff --git a/backend/routes/models.py b/backend/routes/models.py
index c17d0f15..5706c7ad 100644
--- a/backend/routes/models.py
+++ b/backend/routes/models.py
@@ -9,7 +9,7 @@ from fastapi.responses import StreamingResponse
 from sqlalchemy.orm import Session
 
 from .. import models
-from ..platform_detect import get_backend_type
+from ..utils.platform_detect import get_backend_type
 from ..services.task_queue import create_background_task
 from ..utils.progress import get_progress_manager
 from ..utils.tasks import get_task_manager
@@ -50,7 +50,7 @@ def _copy_with_progress(src: Path, dst: Path, progress_manager, copied_so_far: i
 @router.post("/models/load")
 async def load_model(model_size: str = "1.7B"):
     """Manually load TTS model."""
-    from .. import tts
+    from ..services import tts
 
     try:
         tts_model = tts.get_tts_model()
@@ -63,7 +63,7 @@ async def load_model(model_size: str = "1.7B"):
 @router.post("/models/unload")
 async def unload_model():
     """Unload the default Qwen TTS model to free memory."""
-    from .. import tts
+    from ..services import tts
 
     try:
         tts.unload_tts_model()
diff --git a/backend/routes/profiles.py b/backend/routes/profiles.py
index 320dfc50..6b7b4509 100644
--- a/backend/routes/profiles.py
+++ b/backend/routes/profiles.py
@@ -9,10 +9,11 @@ from fastapi import APIRouter, Depends, File, Form, HTTPException, UploadFile
 from fastapi.responses import FileResponse, StreamingResponse
 from sqlalchemy.orm import Session
 
-from .. import channels, config, export_import, models, profiles
+from .. import config, models
 from ..app import safe_content_disposition
 from ..database import VoiceProfile as DBVoiceProfile, get_db
-from ..profiles import _profile_to_response
+from ..services import channels, export_import, profiles
+from ..services.profiles import _profile_to_response
 
 router = APIRouter()
 
diff --git a/backend/routes/stories.py b/backend/routes/stories.py
index 0c5296ba..74af7a50 100644
--- a/backend/routes/stories.py
+++ b/backend/routes/stories.py
@@ -6,7 +6,8 @@ from fastapi import APIRouter, Depends, HTTPException
 from fastapi.responses import StreamingResponse
 from sqlalchemy.orm import Session
 
-from .. import database, models, stories
+from .. import database, models
+from ..services import stories
 from ..app import safe_content_disposition
 from ..database import get_db
 
diff --git a/backend/routes/transcription.py b/backend/routes/transcription.py
index 8edafef6..b5008c82 100644
--- a/backend/routes/transcription.py
+++ b/backend/routes/transcription.py
@@ -6,7 +6,8 @@ from pathlib import Path
 
 from fastapi import APIRouter, File, Form, HTTPException, UploadFile
 
-from .. import models, transcribe
+from .. import models
+from ..services import transcribe
 from ..services.task_queue import create_background_task
 from ..utils.tasks import get_task_manager
 
diff --git a/backend/channels.py b/backend/services/channels.py
similarity index 99%
rename from backend/channels.py
rename to backend/services/channels.py
index 146c003d..f7d9d788 100644
--- a/backend/channels.py
+++ b/backend/services/channels.py
@@ -7,14 +7,14 @@ from datetime import datetime
 import uuid
 from sqlalchemy.orm import Session
 
-from .models import (
+from ..models import (
     AudioChannelCreate,
     AudioChannelUpdate,
     AudioChannelResponse,
     ChannelVoiceAssignment,
     ProfileChannelAssignment,
 )
-from .database import (
+from ..database import (
     AudioChannel as DBAudioChannel,
     ChannelDeviceMapping as DBChannelDeviceMapping,
     ProfileChannelMapping as DBProfileChannelMapping,
diff --git a/backend/effects.py b/backend/services/effects.py
similarity index 96%
rename from backend/effects.py
rename to backend/services/effects.py
index 79130785..29ca5918 100644
--- a/backend/effects.py
+++ b/backend/services/effects.py
@@ -11,8 +11,8 @@ from typing import List, Optional
 from sqlalchemy.orm import Session
 from sqlalchemy.exc import IntegrityError
 
-from .database import EffectPreset as DBEffectPreset
-from .models import EffectPresetResponse, EffectPresetCreate, EffectPresetUpdate, EffectConfig
+from ..database import EffectPreset as DBEffectPreset
+from ..models import EffectPresetResponse, EffectPresetCreate, EffectPresetUpdate, EffectConfig
 
 
 def _preset_response(p: DBEffectPreset) -> EffectPresetResponse:
diff --git a/backend/export_import.py b/backend/services/export_import.py
similarity index 98%
rename from backend/export_import.py
rename to backend/services/export_import.py
index 58acd400..93252f50 100644
--- a/backend/export_import.py
+++ b/backend/services/export_import.py
@@ -12,11 +12,11 @@ from pathlib import Path
 from typing import Optional
 from sqlalchemy.orm import Session
 
-from .models import VoiceProfileResponse
-from .database import VoiceProfile as DBVoiceProfile, ProfileSample as DBProfileSample, Generation as DBGeneration, GenerationVersion as DBGenerationVersion
+from ..models import VoiceProfileResponse
+from ..database import VoiceProfile as DBVoiceProfile, ProfileSample as DBProfileSample, Generation as DBGeneration, GenerationVersion as DBGenerationVersion
 from .profiles import create_profile, add_profile_sample
-from .models import VoiceProfileCreate
-from . import config
+from ..models import VoiceProfileCreate
+from .. import config
 
 
 def _get_unique_profile_name(name: str, db: Session) -> str:
@@ -346,7 +346,7 @@ async def import_generation_from_zip(file_bytes: bytes, db: Session) -> dict:
     import tempfile
     import shutil
     from datetime import datetime
-    from . import config
+    from .. import config
     
     zip_buffer = io.BytesIO(file_bytes)
     
diff --git a/backend/services/generation.py b/backend/services/generation.py
index 6f85cd0c..04e07014 100644
--- a/backend/services/generation.py
+++ b/backend/services/generation.py
@@ -19,7 +19,8 @@ from __future__ import annotations
 import traceback
 from typing import Literal, Optional
 
-from .. import config, history, profiles
+from .. import config
+from . import history, profiles
 from ..database import get_db
 from ..utils.tasks import get_task_manager
 
@@ -151,7 +152,7 @@ def _save_generate(
     Returns the final audio path (processed if effects were applied,
     otherwise clean).
     """
-    from .. import versions as versions_mod
+    from . import versions as versions_mod
 
     clean_audio_path = config.get_generations_dir() / f"{generation_id}.wav"
     save_audio(audio, str(clean_audio_path), sample_rate)
@@ -221,7 +222,7 @@ def _save_regenerate(
 
     Returns the audio path.
     """
-    from .. import versions as versions_mod
+    from . import versions as versions_mod
 
     suffix = version_id[:8] if version_id else generation_id[:8]
     audio_path = config.get_generations_dir() / f"{generation_id}_{suffix}.wav"
diff --git a/backend/history.py b/backend/services/history.py
similarity index 96%
rename from backend/history.py
rename to backend/services/history.py
index c2c9197a..8f45d48f 100644
--- a/backend/history.py
+++ b/backend/services/history.py
@@ -10,9 +10,9 @@ from pathlib import Path
 from sqlalchemy.orm import Session
 from sqlalchemy import or_
 
-from .models import GenerationRequest, GenerationResponse, HistoryQuery, HistoryResponse, HistoryListResponse, GenerationVersionResponse, EffectConfig
-from .database import Generation as DBGeneration, GenerationVersion as DBGenerationVersion, VoiceProfile as DBVoiceProfile
-from . import config
+from ..models import GenerationRequest, GenerationResponse, HistoryQuery, HistoryResponse, HistoryListResponse, GenerationVersionResponse, EffectConfig
+from ..database import Generation as DBGeneration, GenerationVersion as DBGenerationVersion, VoiceProfile as DBVoiceProfile
+from .. import config
 
 
 def _get_versions_for_generation(generation_id: str, db: Session) -> tuple:
diff --git a/backend/profiles.py b/backend/services/profiles.py
similarity index 97%
rename from backend/profiles.py
rename to backend/services/profiles.py
index ab8941c4..46180f3c 100644
--- a/backend/profiles.py
+++ b/backend/services/profiles.py
@@ -10,23 +10,23 @@ from pathlib import Path
 from sqlalchemy.orm import Session
 from sqlalchemy import func, select
 
-from .models import (
+from ..models import (
     VoiceProfileCreate,
     VoiceProfileResponse,
     ProfileSampleCreate,
     ProfileSampleResponse,
 )
-from .database import (
+from ..database import (
     VoiceProfile as DBVoiceProfile,
     ProfileSample as DBProfileSample,
     Generation as DBGeneration,
 )
-from .models import EffectConfig
-from .utils.audio import validate_reference_audio, load_audio, save_audio
-from .utils.images import validate_image, process_avatar
-from .utils.cache import _get_cache_dir, clear_profile_cache
+from ..models import EffectConfig
+from ..utils.audio import validate_reference_audio, load_audio, save_audio
+from ..utils.images import validate_image, process_avatar
+from ..utils.cache import _get_cache_dir, clear_profile_cache
 from .tts import get_tts_model
-from . import config
+from .. import config
 import json as _json
 
 
@@ -389,7 +389,7 @@ async def create_voice_prompt_for_profile(
     Returns:
         Voice prompt dictionary
     """
-    from .backends import get_tts_backend_for_engine
+    from ..backends import get_tts_backend_for_engine
 
     samples = db.query(DBProfileSample).filter_by(profile_id=profile_id).all()
 
diff --git a/backend/stories.py b/backend/services/stories.py
similarity index 99%
rename from backend/stories.py
rename to backend/services/stories.py
index cb4dad65..ac8e22bd 100644
--- a/backend/stories.py
+++ b/backend/services/stories.py
@@ -10,7 +10,7 @@ from pathlib import Path
 from sqlalchemy.orm import Session
 from sqlalchemy import func
 
-from .models import (
+from ..models import (
     StoryCreate,
     StoryResponse,
     StoryDetailResponse,
@@ -22,14 +22,14 @@ from .models import (
     StoryItemSplit,
     StoryItemVersionUpdate,
 )
-from .database import (
+from ..database import (
     Story as DBStory,
     StoryItem as DBStoryItem,
     Generation as DBGeneration,
     VoiceProfile as DBVoiceProfile,
 )
 from .history import _get_versions_for_generation
-from .utils.audio import load_audio, save_audio
+from ..utils.audio import load_audio, save_audio
 import numpy as np
 
 
@@ -754,7 +754,7 @@ async def set_story_item_version(
 
     # Validate version_id belongs to this generation if provided
     if data.version_id:
-        from .database import GenerationVersion as DBGenerationVersion
+        from ..database import GenerationVersion as DBGenerationVersion
 
         version = (
             db.query(DBGenerationVersion)
@@ -820,7 +820,7 @@ async def export_story_audio(
         # Resolve audio path: use pinned version if set, otherwise generation default
         resolved_audio_path = generation.audio_path
         if getattr(item, "version_id", None):
-            from .database import GenerationVersion as DBGenerationVersion
+            from ..database import GenerationVersion as DBGenerationVersion
 
             version = db.query(DBGenerationVersion).filter_by(id=item.version_id).first()
             if version:
diff --git a/backend/transcribe.py b/backend/services/transcribe.py
similarity index 89%
rename from backend/transcribe.py
rename to backend/services/transcribe.py
index 4d735fce..e400dbd5 100644
--- a/backend/transcribe.py
+++ b/backend/services/transcribe.py
@@ -3,7 +3,7 @@ STT (Speech-to-Text) module - delegates to backend abstraction layer.
 """
 
 from typing import Optional
-from .backends import get_stt_backend, STTBackend
+from ..backends import get_stt_backend, STTBackend
 
 
 def get_whisper_model() -> STTBackend:
diff --git a/backend/tts.py b/backend/services/tts.py
similarity index 93%
rename from backend/tts.py
rename to backend/services/tts.py
index 453f2f5c..d4f90ff3 100644
--- a/backend/tts.py
+++ b/backend/services/tts.py
@@ -7,7 +7,7 @@ import numpy as np
 import io
 import soundfile as sf
 
-from .backends import get_tts_backend, TTSBackend
+from ..backends import get_tts_backend, TTSBackend
 
 
 def get_tts_model() -> TTSBackend:
diff --git a/backend/versions.py b/backend/services/versions.py
similarity index 98%
rename from backend/versions.py
rename to backend/services/versions.py
index 381114d7..1743a25c 100644
--- a/backend/versions.py
+++ b/backend/services/versions.py
@@ -14,12 +14,12 @@ from typing import List, Optional
 
 from sqlalchemy.orm import Session
 
-from .database import (
+from ..database import (
     GenerationVersion as DBGenerationVersion,
     Generation as DBGeneration,
 )
-from .models import GenerationVersionResponse, EffectConfig
-from . import config
+from ..models import GenerationVersionResponse, EffectConfig
+from .. import config
 
 
 def _version_response(v: DBGenerationVersion) -> GenerationVersionResponse:
diff --git a/backend/platform_detect.py b/backend/utils/platform_detect.py
similarity index 100%
rename from backend/platform_detect.py
rename to backend/utils/platform_detect.py