diff --git a/backend/configs/config.yaml b/backend/configs/config.yaml
index f0e3e925..044424c6 100644
--- a/backend/configs/config.yaml
+++ b/backend/configs/config.yaml
@@ -3,6 +3,8 @@ whisper:
   model_size: large-v2
   # Compute type. 'float16' for CUDA, 'float32' for CPU.
   compute_type: float16
+  # Whether to offload the model after the inference.
+  enable_offload: true
 
 bgm_separation:
   # UVR model sizes between ["UVR-MDX-NET-Inst_HQ_4", "UVR-MDX-NET-Inst_3"]
diff --git a/configs/default_parameters.yaml b/configs/default_parameters.yaml
index 4a7172fc..022bd984 100644
--- a/configs/default_parameters.yaml
+++ b/configs/default_parameters.yaml
@@ -31,6 +31,7 @@ whisper:
   language_detection_threshold: 0.5
   language_detection_segments: 1
   add_timestamp: false
+  enable_offload: true
 
 vad:
   vad_filter: false
@@ -43,6 +44,7 @@ vad:
 diarization:
   is_diarize: false
   hf_token: ""
+  enable_offload: true
 
 bgm_separation:
   is_separate_bgm: false
diff --git a/configs/translation.yaml b/configs/translation.yaml
index a85124ab..9ea0a90e 100644
--- a/configs/translation.yaml
+++ b/configs/translation.yaml
@@ -19,7 +19,7 @@ en: # English
   Enabling this will remove background music: Enabling this will remove background music by submodel before transcribing
   Enable Background Music Remover Filter: Enable Background Music Remover Filter
   Save separated files to output: Save separated files to output
-  Offload sub model after removing background music: Offload sub model after removing background music
+  Offload sub model when finished: Offload sub model when finished
   Voice Detection Filter: Voice Detection Filter
   Enable this to transcribe only detected voice: Enable this to transcribe only detected voice parts by submodel.
   Enable Silero VAD Filter: Enable Silero VAD Filter
@@ -65,7 +65,7 @@ ko: # Korean
   Enabling this will remove background music: 받아쓰기 이전에 먼저 배경 음악 제거용 서브 모델을 활성화 합니다.
   Enable Background Music Remover Filter: 배경 음악 제거 필터 활성화
   Save separated files to output: 분리된 배경 음악 & 음성 파일 따로 출력 폴더에 저장
-  Offload sub model after removing background music: 배경 음악 제거 후 서브 모델을 비활성화 합니다. (VRAM 이 부족할 시 체크하세요.)
+  Offload sub model when finished: 완료 후 모델 오프로드. (VRAM 이 부족할 시 체크하세요.)
   Voice Detection Filter: 목소리 감지 필터
   Enable this to transcribe only detected voice: 서브 모델에 의해 목소리라고 판단된 부분만 받아쓰기를 진행합니다.
   Enable Silero VAD Filter: Silero VAD 필터 활성화 
@@ -111,7 +111,7 @@ ja: # Japanese
   Enabling this will remove background music: Enabling this will remove background music by submodel before transcribing
   Enable Background Music Remover Filter: Enable Background Music Remover Filter
   Save separated files to output: Save separated files to output
-  Offload sub model after removing background music: Offload sub model after removing background music
+  Offload sub model when finished: Offload sub model when finished
   Voice Detection Filter: Voice Detection Filter
   Enable this to transcribe only detected voice: Enable this to transcribe only detected voice parts by submodel.
   Enable Silero VAD Filter: Enable Silero VAD Filter
@@ -157,7 +157,7 @@ es: # Spanish
   Enabling this will remove background music: Enabling this will remove background music by submodel before transcribing
   Enable Background Music Remover Filter: Enable Background Music Remover Filter
   Save separated files to output: Save separated files to output
-  Offload sub model after removing background music: Offload sub model after removing background music
+  Offload sub model when finished: Offload sub model when finished
   Voice Detection Filter: Voice Detection Filter
   Enable this to transcribe only detected voice: Enable this to transcribe only detected voice parts by submodel.
   Enable Silero VAD Filter: Enable Silero VAD Filter
@@ -203,7 +203,7 @@ fr: # French
   Enabling this will remove background music: L'activation supprimera la musique de fond via un sous-modèle avant la transcription
   Enable Background Music Remover Filter: Activer le filtre de suppression de musique de fond
   Save separated files to output: Sauvegarder les fichiers séparés dans la sortie
-  Offload sub model after removing background music: Décharger le sous-modèle après avoir supprimé la musique de fond
+  Offload sub model when finished: Décharger le sous-modèle une fois terminé
   Voice Detection Filter: Filtre de détection vocale
   Enable this to transcribe only detected voice: Activer pour transcrire uniquement la voix détectée
   Enable Silero VAD Filter: Activer le filtre Silero VAD
@@ -249,7 +249,7 @@ de: # Deutsch
   Enabling this will remove background music: Aktivierung entfernt Hintergrundmusik mithilfe des Submodells vor der Transkription
   Enable Background Music Remover Filter: Hintergrundmusik-Entfernungsfilter aktivieren
   Save separated files to output: Getrennte Dateien in der Ausgabe speichern
-  Offload sub model after removing background music: Submodell nach Entfernung der Hintergrundmusik entladen
+  Offload sub model when finished: Submodell entladen wenn nicht mehr benötigt
   Voice Detection Filter: Sprachfilter
   Enable this to transcribe only detected voice: Aktivieren, um nur erkannte Sprachsegmente mithilfe des Submodells zu transkribieren.
   Enable Silero VAD Filter: Silero VAD-Filter aktivieren
@@ -296,7 +296,7 @@ zh: # Chinese
   Enabling this will remove background music: 启用此功能将会在进行转录前去除背景音乐
   Enable Background Music Remover Filter: 去除背景音乐
   Save separated files to output: 导出分离出的音频文件
-  Offload sub model after removing background music: 移除背景音乐后卸载子模型
+  Offload sub model when finished: 完成后卸载子模型
   Voice Detection Filter: 话音检测设置
   Enable this to transcribe only detected voice: 启用此功能将仅转录检测到的语音部分
   Enable Silero VAD Filter: 启用 Silero 语音活动检测 (VAD) 
@@ -342,7 +342,7 @@ uk: # Ukrainian
   Enabling this will remove background music: Увімкнення цього видалить фонову музику за допомогою підмоделі перед транскрипцією
   Enable Background Music Remover Filter: Увімкнути фільтр видалення фонової музики
   Save separated files to output: Зберегти розділені файли до вихідної папки
-  Offload sub model after removing background music: Вивантажити підмодель після видалення фонової музики
+  Offload sub model when finished: Після завершення вивантажте підмодель
   Voice Detection Filter: Фільтр розпізнавання голосу
   Enable this to transcribe only detected voice: Увімкніть це, щоб транскрибувати лише розпізнані голосові частини за допомогою підмоделі
   Enable Silero VAD Filter: Увімкнути фільтр Silero VAD
@@ -388,7 +388,7 @@ ru: # Russian
   Enabling this will remove background music: Включение этого удалит фоновую музыку с помощью подмодели перед транскрипцией
   Enable Background Music Remover Filter: Включить фильтр удаления фоновой музыки
   Save separated files to output: Сохранить разделенные файлы в выходную папку
-  Offload sub model after removing background music: Выгрузить подмодель после удаления фоновой музыки
+  Offload sub model when finished: Выгрузить подмодель после завершения
   Voice Detection Filter: Фильтр обнаружения голоса
   Enable this to transcribe only detected voice: Включите это, чтобы транскрибировать только обнаруженные голосовые части с помощью подмодели
   Enable Silero VAD Filter: Включить фильтр Silero VAD
@@ -434,7 +434,7 @@ tr: # Turkish
   Enabling this will remove background music: Bunu etkinleştirmek, arka plan müziğini alt model tarafından transkripsiyondan önce kaldıracaktır
   Enable Background Music Remover Filter: Arka Plan Müziği Kaldırma Filtresini Etkinleştir
   Save separated files to output: Ayrılmış dosyaları çıktıya kaydet
-  Offload sub model after removing background music: Arka plan müziği kaldırıldıktan sonra alt modeli devre dışı bırak
+  Offload sub model when finished: Alt modeli bitirdiğinizde boşaltın
   Voice Detection Filter: Ses Algılama Filtresi
   Enable this to transcribe only detected voice: Bunu etkinleştirerek yalnızca alt model tarafından algılanan ses kısımlarını transkribe et
   Enable Silero VAD Filter: Silero VAD Filtresini Etkinleştir
@@ -480,7 +480,7 @@ eu: # Basque
   Enabling this will remove background music: Hau aktibatuz, transkribatu aurretik atzeko musika ezabatuko zaio azpieredu baten bidez
   Enable Background Music Remover Filter: Aktibatu Atzeko Musika Ezabatzeko Filtroa
   Save separated files to output: Gorde fitxategi bereiziak irteeran
-  Offload sub model after removing background music: Atzeko musika kendu ondoren azpieredua memoriatik kendu
+  Offload sub model when finished: Deskargatu azpieredua amaitutakoan
   Voice Detection Filter: Ahots Detekzio Filtroa
   Enable this to transcribe only detected voice: Aktibatu hau azpieredu batekin soilik detektatutako ahots zatiak transkribatzeko.
   Enable Silero VAD Filter: Aktibatu Silero VAD Filtroa
diff --git a/modules/whisper/base_transcription_pipeline.py b/modules/whisper/base_transcription_pipeline.py
index 1cd47077..d2fb97ac 100644
--- a/modules/whisper/base_transcription_pipeline.py
+++ b/modules/whisper/base_transcription_pipeline.py
@@ -169,6 +169,8 @@ def run(self,
             progress,
             *whisper_params.to_list()
         )
+        if whisper_params.enable_offload:
+            self.offload()
 
         if vad_params.vad_filter:
             restored_result = self.vad.restore_speech_timestamps(
@@ -188,6 +190,8 @@ def run(self,
                 transcribed_result=result,
                 device=diarization_params.diarization_device
             )
+            if diarization_params.enable_offload:
+                self.diarizer.offload()
 
         self.cache_parameters(
             params=params,
diff --git a/modules/whisper/data_classes.py b/modules/whisper/data_classes.py
index ad72ee33..700e2a8a 100644
--- a/modules/whisper/data_classes.py
+++ b/modules/whisper/data_classes.py
@@ -161,6 +161,10 @@ class DiarizationParams(BaseParams):
         default="",
         description="Hugging Face token for downloading diarization models"
     )
+    enable_offload: bool = Field(
+        default=True,
+        description="Offload Diarization model after Speaker diarization"
+    )
 
     @classmethod
     def to_gradio_inputs(cls,
@@ -182,6 +186,10 @@ def to_gradio_inputs(cls,
                 value=defaults.get("hf_token", cls.__fields__["hf_token"].default),
                 info=_("This is only needed the first time you download the model")
             ),
+            gr.Checkbox(
+                label=_("Offload sub model when finished"),
+                value=defaults.get("enable_offload", cls.__fields__["enable_offload"].default),
+            )
         ]
 
 
@@ -242,7 +250,7 @@ def to_gradio_input(cls,
                 value=defaults.get("save_file", cls.__fields__["save_file"].default),
             ),
             gr.Checkbox(
-                label=_("Offload sub model after removing background music"),
+                label=_("Offload sub model when finished"),
                 value=defaults.get("enable_offload", cls.__fields__["enable_offload"].default),
             )
         ]
@@ -328,6 +336,10 @@ class WhisperParams(BaseParams):
         description="Number of segments for language detection"
     )
     batch_size: int = Field(default=24, gt=0, description="Batch size for processing")
+    enable_offload: bool = Field(
+        default=True,
+        description="Offload Whisper model after transcription"
+    )
 
     @field_validator('lang')
     def validate_lang(cls, v):
@@ -556,6 +568,13 @@ def to_gradio_inputs(cls,
 
         inputs += faster_whisper_inputs + insanely_fast_whisper_inputs
 
+        inputs += [
+            gr.Checkbox(
+                label=_("Offload sub model when finished"),
+                value=defaults.get("enable_offload", cls.__fields__["enable_offload"].default),
+            )
+        ]
+
         return inputs