diff --git a/backend/backends/base.py b/backend/backends/base.py
index 70ec11ef3..d5fd8532f 100644
--- a/backend/backends/base.py
+++ b/backend/backends/base.py
@@ -171,17 +171,23 @@ def check_cuda_compatibility() -> tuple[bool, str | None]:
 
 def empty_device_cache(device: str) -> None:
     """
-    Free cached memory on the given device (CUDA or XPU).
+    Free cached memory and unreferenced tensors on the given device (CUDA, XPU, MPS, CPU).
 
-    Backends should call this after unloading models so VRAM is returned
-    to the OS.
+    Backends call this after model unloading and post-generation cleanup to return
+    memory to the OS and prevent process heap accumulation.
     """
+    import gc
     import torch
 
+    gc.collect()
+
     if device == "cuda" and torch.cuda.is_available():
         torch.cuda.empty_cache()
     elif device == "xpu" and hasattr(torch, "xpu"):
         torch.xpu.empty_cache()
+    elif device == "mps" and hasattr(torch.backends, "mps") and torch.backends.mps.is_available():
+        if hasattr(torch.mps, "empty_cache"):
+            torch.mps.empty_cache()
 
 
 def manual_seed(seed: int, device: str) -> None:
diff --git a/backend/backends/chatterbox_backend.py b/backend/backends/chatterbox_backend.py
index e7a025b38..8ea2172dd 100644
--- a/backend/backends/chatterbox_backend.py
+++ b/backend/backends/chatterbox_backend.py
@@ -203,21 +203,22 @@ def _generate_sync():
 
             logger.info(f"[Chatterbox] Generating: lang={language}")
 
-            wav = self.model.generate(
-                text,
-                language_id=language,
-                audio_prompt_path=ref_audio,
-                exaggeration=lang_defaults["exaggeration"],
-                cfg_weight=lang_defaults["cfg_weight"],
-                temperature=lang_defaults["temperature"],
-                repetition_penalty=lang_defaults["repetition_penalty"],
-            )
-
-            # Convert tensor -> numpy
-            if isinstance(wav, torch.Tensor):
-                audio = wav.squeeze().cpu().numpy().astype(np.float32)
-            else:
-                audio = np.asarray(wav, dtype=np.float32)
+            with torch.inference_mode():
+                wav = self.model.generate(
+                    text,
+                    language_id=language,
+                    audio_prompt_path=ref_audio,
+                    exaggeration=lang_defaults["exaggeration"],
+                    cfg_weight=lang_defaults["cfg_weight"],
+                    temperature=lang_defaults["temperature"],
+                    repetition_penalty=lang_defaults["repetition_penalty"],
+                )
+
+                # Convert tensor -> numpy
+                if isinstance(wav, torch.Tensor):
+                    audio = wav.squeeze().cpu().numpy().astype(np.float32)
+                else:
+                    audio = np.asarray(wav, dtype=np.float32)
 
             sample_rate = getattr(self.model, "sr", None) or getattr(self.model, "sample_rate", 24000)
 
diff --git a/backend/backends/chatterbox_turbo_backend.py b/backend/backends/chatterbox_turbo_backend.py
index 6f7d6b94f..29be69e8d 100644
--- a/backend/backends/chatterbox_turbo_backend.py
+++ b/backend/backends/chatterbox_turbo_backend.py
@@ -184,20 +184,21 @@ def _generate_sync():
 
             logger.info("[Chatterbox Turbo] Generating (English)")
 
-            wav = self.model.generate(
-                text,
-                audio_prompt_path=ref_audio,
-                temperature=0.8,
-                top_k=1000,
-                top_p=0.95,
-                repetition_penalty=1.2,
-            )
-
-            # Convert tensor -> numpy
-            if isinstance(wav, torch.Tensor):
-                audio = wav.squeeze().cpu().numpy().astype(np.float32)
-            else:
-                audio = np.asarray(wav, dtype=np.float32)
+            with torch.inference_mode():
+                wav = self.model.generate(
+                    text,
+                    audio_prompt_path=ref_audio,
+                    temperature=0.8,
+                    top_k=1000,
+                    top_p=0.95,
+                    repetition_penalty=1.2,
+                )
+
+                # Convert tensor -> numpy
+                if isinstance(wav, torch.Tensor):
+                    audio = wav.squeeze().cpu().numpy().astype(np.float32)
+                else:
+                    audio = np.asarray(wav, dtype=np.float32)
 
             sample_rate = getattr(self.model, "sr", None) or getattr(self.model, "sample_rate", 24000)
 
diff --git a/backend/backends/kokoro_backend.py b/backend/backends/kokoro_backend.py
index 59005f0a7..0f6ff39ec 100644
--- a/backend/backends/kokoro_backend.py
+++ b/backend/backends/kokoro_backend.py
@@ -276,12 +276,13 @@ def _generate_sync():
 
             # Generate all chunks and concatenate
             audio_chunks = []
-            for result in pipeline(text, voice=voice_name, speed=1.0):
-                if result.audio is not None:
-                    chunk = result.audio
-                    if isinstance(chunk, torch.Tensor):
-                        chunk = chunk.detach().cpu().numpy()
-                    audio_chunks.append(chunk.squeeze())
+            with torch.inference_mode():
+                for result in pipeline(text, voice=voice_name, speed=1.0):
+                    if result.audio is not None:
+                        chunk = result.audio
+                        if isinstance(chunk, torch.Tensor):
+                            chunk = chunk.detach().cpu().numpy()
+                        audio_chunks.append(chunk.squeeze())
 
             if not audio_chunks:
                 # Return 1 second of silence as fallback
diff --git a/backend/backends/luxtts_backend.py b/backend/backends/luxtts_backend.py
index 7f15686af..55be18b24 100644
--- a/backend/backends/luxtts_backend.py
+++ b/backend/backends/luxtts_backend.py
@@ -167,18 +167,21 @@ def _generate_sync():
             if seed is not None:
                 manual_seed(seed, self.device)
 
-            wav = self.model.generate_speech(
-                text=text,
-                encode_dict=voice_prompt,
-                num_steps=4,
-                guidance_scale=3.0,
-                t_shift=0.5,
-                speed=1.0,
-                return_smooth=False,  # 48kHz output
-            )
+            import torch
+
+            with torch.inference_mode():
+                wav = self.model.generate_speech(
+                    text=text,
+                    encode_dict=voice_prompt,
+                    num_steps=4,
+                    guidance_scale=3.0,
+                    t_shift=0.5,
+                    speed=1.0,
+                    return_smooth=False,  # 48kHz output
+                )
 
-            # LuxTTS returns a tensor (may be on GPU/MPS), move to CPU first
-            audio = wav.detach().cpu().numpy().squeeze()
+                # LuxTTS returns a tensor (may be on GPU/MPS), move to CPU first
+                audio = wav.detach().cpu().numpy().squeeze()
             return audio, 48000
 
         return await asyncio.to_thread(_generate_sync)
diff --git a/backend/backends/pytorch_backend.py b/backend/backends/pytorch_backend.py
index f8ae79b86..faca0d6ab 100644
--- a/backend/backends/pytorch_backend.py
+++ b/backend/backends/pytorch_backend.py
@@ -231,12 +231,13 @@ def _generate_sync():
 
             # See _create_prompt_sync comment — inference runs with the
             # process's default HF_HUB_OFFLINE state (issue #462).
-            wavs, sample_rate = self.model.generate_voice_clone(
-                text=text,
-                voice_clone_prompt=voice_prompt,
-                language=LANGUAGE_CODE_TO_NAME.get(language, "auto"),
-                instruct=instruct,
-            )
+            with torch.inference_mode():
+                wavs, sample_rate = self.model.generate_voice_clone(
+                    text=text,
+                    voice_clone_prompt=voice_prompt,
+                    language=LANGUAGE_CODE_TO_NAME.get(language, "auto"),
+                    instruct=instruct,
+                )
             return wavs[0], sample_rate
 
         # Run blocking inference in thread pool to avoid blocking event loop
diff --git a/backend/backends/qwen_custom_voice_backend.py b/backend/backends/qwen_custom_voice_backend.py
index 74f739bba..791f800b0 100644
--- a/backend/backends/qwen_custom_voice_backend.py
+++ b/backend/backends/qwen_custom_voice_backend.py
@@ -207,7 +207,8 @@ def _generate_sync():
             # state. Forcing offline here (issue #462) regressed online
             # users whose libraries issue legitimate metadata lookups
             # during generation.
-            wavs, sample_rate = self.model.generate_custom_voice(**kwargs)
+            with torch.inference_mode():
+                wavs, sample_rate = self.model.generate_custom_voice(**kwargs)
             return wavs[0], sample_rate
 
         audio, sample_rate = await asyncio.to_thread(_generate_sync)
diff --git a/backend/services/generation.py b/backend/services/generation.py
index a4b2e8a3f..e1893d8d2 100644
--- a/backend/services/generation.py
+++ b/backend/services/generation.py
@@ -156,6 +156,12 @@ async def run_generation(
     finally:
         task_manager.complete_generation(generation_id)
         bg_db.close()
+        try:
+            from ..backends.base import empty_device_cache
+            device = getattr(tts_model, "device", "cpu") if "tts_model" in locals() else "cpu"
+            empty_device_cache(device)
+        except Exception:
+            pass
 
 
 def _notify_speak_end(generation_id: str, *, status: str) -> None:
@@ -313,14 +319,22 @@ async def generate_audio_sync(
     if crossfade_ms is not None:
         gen_kwargs["crossfade_ms"] = crossfade_ms
 
-    audio, sample_rate = await generate_chunked(
-        tts_model, text, voice_prompt, **gen_kwargs
-    )
+    try:
+        audio, sample_rate = await generate_chunked(
+            tts_model, text, voice_prompt, **gen_kwargs
+        )
 
-    if normalize:
-        audio = normalize_audio(audio)
+        if normalize:
+            audio = normalize_audio(audio)
 
-    return tts.audio_to_wav_bytes(audio, sample_rate)
+        return tts.audio_to_wav_bytes(audio, sample_rate)
+    finally:
+        try:
+            from ..backends.base import empty_device_cache
+            device = getattr(tts_model, "device", "cpu") if "tts_model" in locals() else "cpu"
+            empty_device_cache(device)
+        except Exception:
+            pass
 
 
 def _save_regenerate(