diff --git a/src/openbench/cli/commands/evaluate.py b/src/openbench/cli/commands/evaluate.py
index 3569da8..1b4f5f7 100644
--- a/src/openbench/cli/commands/evaluate.py
+++ b/src/openbench/cli/commands/evaluate.py
@@ -175,6 +175,7 @@ def run_alias_mode(
     wandb_run_name: str | None,
     wandb_tags: list[str] | None,
     use_keywords: bool | None,
+    force_language: bool,
     verbose: bool,
 ) -> BenchmarkResult:
     """Run evaluation using pipeline and dataset aliases."""
@@ -201,6 +202,12 @@ def run_alias_mode(
             if verbose:
                 typer.echo(f"✅ Keywords: {'enabled' if use_keywords else 'disabled'} (override)")
 
+        # Handle force_language override
+        if force_language:
+            pipeline_config_override["force_language"] = force_language
+            if verbose:
+                typer.echo("✅ Force language: enabled")
+
         pipeline = PipelineRegistry.create_pipeline(pipeline_name, config=pipeline_config_override)
 
         ######### Build Benchmark Config #########
@@ -333,6 +340,11 @@ def evaluate(
         "--use-keywords",
         help="Enable keyword boosting for compatible pipelines (overrides default config)",
     ),
+    force_language: bool = typer.Option(
+        False,
+        "--force-language",
+        help="Force language hinting for compatible pipelines",
+    ),
     verbose: bool = typer.Option(False, "--verbose", "-v", help="Enable verbose output"),
 ) -> None:
     """Run evaluation benchmarks.
@@ -393,6 +405,7 @@ def evaluate(
                 wandb_run_name=wandb_run_name,
                 wandb_tags=wandb_tags,
                 use_keywords=use_keywords,
+                force_language=force_language,
                 verbose=verbose,
             )
         display_result(result)
diff --git a/src/openbench/dataset/dataset_aliases.py b/src/openbench/dataset/dataset_aliases.py
index 5d20a2a..0787913 100644
--- a/src/openbench/dataset/dataset_aliases.py
+++ b/src/openbench/dataset/dataset_aliases.py
@@ -344,7 +344,11 @@ def register_dataset_aliases() -> None:
     # Portuguese
     DatasetRegistry.register_alias(
         "common-voice-pt",
-        DatasetConfig(dataset_id="argmaxinc/common_voice_17_0-argmax_subset-400-openbench", split="test", subset="pt"),
+        DatasetConfig(
+            dataset_id="argmaxinc/common_voice_17_0-argmax_subset-400-openbench",
+            split="test",
+            subset="pt",
+        ),
         supported_pipeline_types={
             PipelineType.TRANSCRIPTION,
         },
diff --git a/src/openbench/dataset/dataset_orchestration.py b/src/openbench/dataset/dataset_orchestration.py
index a12d7d4..0ee3dda 100644
--- a/src/openbench/dataset/dataset_orchestration.py
+++ b/src/openbench/dataset/dataset_orchestration.py
@@ -1,8 +1,6 @@
 # For licensing see accompanying LICENSE.md file.
 # Copyright (C) 2025 Argmax, Inc. All Rights Reserved.
 
-from typing import Any
-
 from datasets import Audio as HfAudio
 from pydantic import model_validator
 from typing_extensions import NotRequired, TypedDict
@@ -11,6 +9,12 @@
 from .dataset_base import BaseDataset, BaseSample
 
 
+class OrchestrationExtraInfo(TypedDict, total=False):
+    """Extra info for orchestration samples."""
+
+    language: str
+
+
 class OrchestrationRow(TypedDict):
     """Expected row structure for orchestration datasets."""
 
@@ -19,9 +23,10 @@ class OrchestrationRow(TypedDict):
     word_speakers: list[str]
     word_timestamps_start: NotRequired[list[float]]
     word_timestamps_end: NotRequired[list[float]]
+    language: NotRequired[str]
 
 
-class OrchestrationSample(BaseSample[Transcript, dict[str, Any]]):
+class OrchestrationSample(BaseSample[Transcript, OrchestrationExtraInfo]):
     """Orchestration sample with speaker validation."""
 
     @model_validator(mode="after")
@@ -31,6 +36,11 @@ def validate_speaker_labels(self) -> "OrchestrationSample":
             raise ValueError("Orchestration samples require transcript with speaker labels")
         return self
 
+    @property
+    def language(self) -> str | None:
+        """Convenience property to access language from extra_info."""
+        return self.extra_info.get("language")
+
 
 class OrchestrationDataset(BaseDataset[OrchestrationSample]):
     """Dataset for orchestration pipelines."""
@@ -38,7 +48,7 @@ class OrchestrationDataset(BaseDataset[OrchestrationSample]):
     _expected_columns = ["audio", "transcript", "word_speakers"]
     _sample_class = OrchestrationSample
 
-    def prepare_sample(self, row: OrchestrationRow) -> tuple[Transcript, dict[str, Any]]:
+    def prepare_sample(self, row: OrchestrationRow) -> tuple[Transcript, OrchestrationExtraInfo]:
         """Prepare transcript with speaker labels and extra info from dataset row."""
         transcript_words = row["transcript"]
         word_speakers = row["word_speakers"]
@@ -52,5 +62,7 @@ def prepare_sample(self, row: OrchestrationRow) -> tuple[Transcript, dict[str, A
             end=row.get("word_timestamps_end"),
             speaker=word_speakers,
         )
-        extra_info: dict[str, Any] = {}
+        extra_info: OrchestrationExtraInfo = {}
+        if "language" in row:
+            extra_info["language"] = row["language"]
         return reference, extra_info
diff --git a/src/openbench/engine/deepgram_engine.py b/src/openbench/engine/deepgram_engine.py
index da75286..1da0745 100644
--- a/src/openbench/engine/deepgram_engine.py
+++ b/src/openbench/engine/deepgram_engine.py
@@ -38,6 +38,9 @@ def __init__(self, options: PrerecordedOptions, timeout: Timeout = Timeout(300))
 
         self.client = DeepgramClient(os.getenv("DEEPGRAM_API_KEY"))
 
+    def set_language(self, language: str) -> None:
+        self.options.language = language
+
     # Only intended to be used with offiline transcription
     def transcribe(self, audio_path: Path | str, keyterm: str | None = None) -> DeepgramApiResponse:
         # Manually construct URL with keyterm parameter using + separator
diff --git a/src/openbench/engine/openai_engine.py b/src/openbench/engine/openai_engine.py
index 9f24ddd..258fa00 100644
--- a/src/openbench/engine/openai_engine.py
+++ b/src/openbench/engine/openai_engine.py
@@ -37,7 +37,7 @@ def get_transcription_kwargs(self) -> dict[str, Any]:
         }
 
     def transcribe(
-        self, audio_path: Path | str, prompt: str | None = None
+        self, audio_path: Path | str, prompt: str | None = None, language: str | None = None
     ) -> TranscriptionVerbose | TranscriptionDiarized:
         if isinstance(audio_path, str):
             audio_path = Path(audio_path)
@@ -50,6 +50,9 @@ def transcribe(
             if prompt is not None:
                 kwargs["prompt"] = prompt
 
+            if language is not None:
+                kwargs["language"] = language
+
             response = self.client.audio.transcriptions.create(**kwargs)
 
         return response
diff --git a/src/openbench/engine/whisperkitpro_engine.py b/src/openbench/engine/whisperkitpro_engine.py
index bd56fcd..8a855e9 100644
--- a/src/openbench/engine/whisperkitpro_engine.py
+++ b/src/openbench/engine/whisperkitpro_engine.py
@@ -227,7 +227,6 @@ def download_and_prepare_model(self) -> Path:
                 raise RuntimeError(f"Model download succeeded but path doesn't exist: {model_path}")
 
             return model_path
-
         except Exception as e:
             raise RuntimeError(f"Failed to download model from {self.repo_id}: {e}") from e
 
@@ -238,6 +237,7 @@ class WhisperKitProInput(BaseModel):
     audio_path: Path
     keep_audio: bool = False
     custom_vocabulary_path: str | None = Field(None, description="Optional path to custom vocabulary file")
+    language: str | None = Field(None, description="Optional language hint for transcription")
 
 
 class WhisperKitProOutput(BaseModel):
@@ -295,6 +295,10 @@ def __call__(self, input: WhisperKitProInput) -> WhisperKitProOutput:
         if input.custom_vocabulary_path:
             cmd.extend(["--custom-vocabulary-path", input.custom_vocabulary_path])
 
+        # Add language hint if provided
+        if input.language:
+            cmd.extend(["--language", input.language])
+
         if "WHISPERKITPRO_API_KEY" in os.environ:
             cmd.extend(["--api-key", os.environ["WHISPERKITPRO_API_KEY"]])
         else:
diff --git a/src/openbench/pipeline/orchestration/common.py b/src/openbench/pipeline/orchestration/common.py
index 5d94b7e..5e2820d 100644
--- a/src/openbench/pipeline/orchestration/common.py
+++ b/src/openbench/pipeline/orchestration/common.py
@@ -23,6 +23,15 @@
 logger = get_logger(__name__)
 
 
+class OrchestrationConfig(PipelineConfig):
+    """Base configuration for orchestration pipelines."""
+
+    force_language: bool = Field(
+        False,
+        description="Force the language of the audio files i.e. hint the model to use the correct language.",
+    )
+
+
 class OrchestrationOutput(PipelineOutput[Transcript]):
     transcription_output: TranscriptionOutput | None = Field(
         default=None,
diff --git a/src/openbench/pipeline/orchestration/nemo/orchestration_mt_parakeet.py b/src/openbench/pipeline/orchestration/nemo/orchestration_mt_parakeet.py
index d352ae6..e656fc9 100644
--- a/src/openbench/pipeline/orchestration/nemo/orchestration_mt_parakeet.py
+++ b/src/openbench/pipeline/orchestration/nemo/orchestration_mt_parakeet.py
@@ -5,7 +5,7 @@
 from typing import Callable
 
 import torch
-from argmaxtools.utils import get_fastest_device
+from argmaxtools.utils import get_fastest_device, get_logger
 from nemo.collections.asr.models import ASRModel, SortformerEncLabelModel
 
 # Use the helper class `SpeakerTaggedASR`, which handles all ASR and diarization cache data for streaming.
@@ -16,22 +16,25 @@
 
 from ....dataset import OrchestrationSample
 from ....pipeline_prediction import Transcript, Word
-from ...base import Pipeline, PipelineConfig, PipelineType, register_pipeline
+from ...base import Pipeline, PipelineType, register_pipeline
 from ...diarization.nemo.sortformer_pipeline import NeMoSortformerPipelineInput
-from ..common import OrchestrationOutput
+from ..common import OrchestrationConfig, OrchestrationOutput
 
 # Use the pre-defined dataclass template `MultitalkerTranscriptionConfig` from `multitalker_transcript_config.py`.
 # Configure the diarization model using streaming parameters:
 from .multitalker_transcript_config import MultitalkerTranscriptionConfig
 
 
+logger = get_logger(__name__)
+
+
 # Constants
 TEMP_AUDIO_DIR = Path("./temp_audio")
 
 __all__ = ["NeMoMTParakeetPipeline", "NeMoMTParakeetPipelineConfig"]
 
 
-class NeMoMTParakeetPipelineConfig(PipelineConfig):
+class NeMoMTParakeetPipelineConfig(OrchestrationConfig):
     diar_model_id: str = Field(
         default="nvidia/diar_streaming_sortformer_4spk-v2.1",
         description="The ID of the diarization model to use.",
@@ -97,6 +100,14 @@ def inference(sample: NeMoSortformerPipelineInput) -> list[dict[str, str]]:
 
     def parse_input(self, input_sample: OrchestrationSample) -> OrchestrationSample:
         assert input_sample.sample_rate == 16000, "Sample rate must be 16kHz"
+
+        # Warn if force_language is enabled (not currently supported)
+        if self.config.force_language:
+            logger.warning(
+                f"{self.__class__.__name__} does not support language hinting. "
+                "The force_language flag will be ignored."
+            )
+
         parsed_input = NeMoSortformerPipelineInput(
             audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
             keep_audio=False,
diff --git a/src/openbench/pipeline/orchestration/orchestration_deepgram.py b/src/openbench/pipeline/orchestration/orchestration_deepgram.py
index 2ac5518..3a95404 100644
--- a/src/openbench/pipeline/orchestration/orchestration_deepgram.py
+++ b/src/openbench/pipeline/orchestration/orchestration_deepgram.py
@@ -1,21 +1,24 @@
 from pathlib import Path
 from typing import Callable
 
+from argmaxtools.utils import get_logger
 from deepgram import PrerecordedOptions
 from pydantic import Field
 
 from ...dataset import DiarizationSample
 from ...engine import DeepgramApi, DeepgramApiResponse
-from ...pipeline import Pipeline, PipelineConfig, register_pipeline
+from ...pipeline import Pipeline, register_pipeline
 from ...pipeline_prediction import Transcript
 from ...types import PipelineType
-from .common import OrchestrationOutput
+from .common import OrchestrationConfig, OrchestrationOutput
 
 
+logger = get_logger(__name__)
+
 TEMP_AUDIO_DIR = Path("temp_audio_dir")
 
 
-class DeepgramOrchestrationPipelineConfig(PipelineConfig):
+class DeepgramOrchestrationPipelineConfig(OrchestrationConfig):
     model_version: str = Field(
         default="nova-3",
         description="The version of the Deepgram model to use",
@@ -28,14 +31,22 @@ class DeepgramOrchestrationPipeline(Pipeline):
     pipeline_type = PipelineType.ORCHESTRATION
 
     def build_pipeline(self) -> Callable[[Path], DeepgramApiResponse]:
-        deepgram_api = DeepgramApi(
+        # Create base API with auto language detection
+        base_api = DeepgramApi(
             options=PrerecordedOptions(
-                model=self.config.model_version, smart_format=True, diarize=True, detect_language=True
+                model=self.config.model_version,
+                smart_format=True,
+                diarize=True,
+                detect_language=not self.config.force_language,
             )
         )
 
         def transcribe(audio_path: Path) -> DeepgramApiResponse:
-            response = deepgram_api.transcribe(audio_path)
+            # Use language-specific API if language is set, otherwise use base API
+            if self.current_language:
+                base_api.set_language(self.current_language)
+
+            response = base_api.transcribe(audio_path)
             # Remove temporary audio path
             audio_path.unlink(missing_ok=True)
             return response
@@ -43,6 +54,11 @@ def transcribe(audio_path: Path) -> DeepgramApiResponse:
         return transcribe
 
     def parse_input(self, input_sample: DiarizationSample) -> Path:
+        # Extract language if force_language is enabled
+        self.current_language = None
+        if self.config.force_language:
+            self.current_language = input_sample.language
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: DeepgramApiResponse) -> OrchestrationOutput:
diff --git a/src/openbench/pipeline/orchestration/orchestration_openai.py b/src/openbench/pipeline/orchestration/orchestration_openai.py
index 9a3fb32..867a67d 100644
--- a/src/openbench/pipeline/orchestration/orchestration_openai.py
+++ b/src/openbench/pipeline/orchestration/orchestration_openai.py
@@ -1,21 +1,24 @@
 from pathlib import Path
 from typing import Callable, Literal
 
+from argmaxtools.utils import get_logger
 from openai.types.audio import TranscriptionDiarized
 from pydantic import Field
 
 from ...dataset import OrchestrationSample
 from ...engine import OpenAIApi
-from ...pipeline import Pipeline, PipelineConfig, register_pipeline
+from ...pipeline import Pipeline, register_pipeline
 from ...pipeline_prediction import Transcript, Word
 from ...types import PipelineType
-from .common import OrchestrationOutput
+from .common import OrchestrationConfig, OrchestrationOutput
 
 
+logger = get_logger(__name__)
+
 TEMP_AUDIO_DIR = Path("temp_audio_dir")
 
 
-class OpenAIOrchestrationPipelineConfig(PipelineConfig):
+class OpenAIOrchestrationPipelineConfig(OrchestrationConfig):
     model_version: Literal["gpt-4o-transcribe-diarize"] = Field(
         default="gpt-4o-transcribe-diarize",
         description="The version of the OpenAI model to use. Currently only `gpt-4o-transcribe-diarize` is supported.",
@@ -31,7 +34,7 @@ def build_pipeline(self) -> Callable[[Path], TranscriptionDiarized]:
         openai_api = OpenAIApi(model=self.config.model_version)
 
         def orchestrate(audio_path: Path) -> TranscriptionDiarized:
-            response = openai_api.transcribe(audio_path)
+            response = openai_api.transcribe(audio_path, language=self.current_language)
             # Remove temporary audio path
             audio_path.unlink(missing_ok=True)
             return response
@@ -39,6 +42,11 @@ def orchestrate(audio_path: Path) -> TranscriptionDiarized:
         return orchestrate
 
     def parse_input(self, input_sample: OrchestrationSample) -> Path:
+        # Extract language if force_language is enabled
+        self.current_language = None
+        if self.config.force_language:
+            self.current_language = input_sample.language
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: TranscriptionDiarized) -> OrchestrationOutput:
diff --git a/src/openbench/pipeline/orchestration/orchestration_whisperkitpro.py b/src/openbench/pipeline/orchestration/orchestration_whisperkitpro.py
index 52fa7b1..577b229 100644
--- a/src/openbench/pipeline/orchestration/orchestration_whisperkitpro.py
+++ b/src/openbench/pipeline/orchestration/orchestration_whisperkitpro.py
@@ -11,8 +11,8 @@
 from ...dataset import OrchestrationSample
 from ...engine import WhisperKitPro, WhisperKitProConfig, WhisperKitProInput, WhisperKitProOutput
 from ...pipeline_prediction import Transcript, Word
-from ..base import Pipeline, PipelineConfig, PipelineType, register_pipeline
-from .common import OrchestrationOutput
+from ..base import Pipeline, PipelineType, register_pipeline
+from .common import OrchestrationConfig, OrchestrationOutput
 
 
 logger = get_logger(__name__)
@@ -20,7 +20,7 @@
 TEMP_AUDIO_DIR = Path("./temp_audio")
 
 
-class WhisperKitProOrchestrationConfig(PipelineConfig):
+class WhisperKitProOrchestrationConfig(OrchestrationConfig):
     cli_path: str = Field(
         ...,
         description="The path to the WhisperKitPro CLI",
@@ -93,9 +93,15 @@ def build_pipeline(self) -> WhisperKitPro:
         return engine
 
     def parse_input(self, input_sample: OrchestrationSample) -> WhisperKitProInput:
+        # Extract language if force_language is enabled
+        language = None
+        if self.config.force_language:
+            language = input_sample.language
+
         return WhisperKitProInput(
             audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
             keep_audio=False,
+            language=language,
         )
 
     def parse_output(self, output: WhisperKitProOutput) -> OrchestrationOutput:
diff --git a/src/openbench/pipeline/orchestration/whisperx.py b/src/openbench/pipeline/orchestration/whisperx.py
index b73513d..498ca34 100644
--- a/src/openbench/pipeline/orchestration/whisperx.py
+++ b/src/openbench/pipeline/orchestration/whisperx.py
@@ -11,8 +11,8 @@
 
 from ...dataset import DiarizationSample
 from ...pipeline_prediction import Transcript
-from ..base import Pipeline, PipelineConfig, PipelineType, register_pipeline
-from .common import OrchestrationOutput
+from ..base import Pipeline, PipelineType, register_pipeline
+from .common import OrchestrationConfig, OrchestrationOutput
 
 
 logger = get_logger(__name__)
@@ -20,7 +20,7 @@
 TEMP_AUDIO_DIR = Path("audio_temp")
 
 
-class WhisperXPipelineConfig(PipelineConfig):
+class WhisperXPipelineConfig(OrchestrationConfig):
     model_name: str = Field(
         default="tiny",
         description="The name of the Whisper model to use",
@@ -43,11 +43,20 @@ class WhisperXPipelineConfig(PipelineConfig):
     )
 
 
+class WhisperXInput:
+    """Input for WhisperX CLI."""
+
+    def __init__(self, audio_path: Path, language: str | None = None):
+        self.audio_path = audio_path
+        self.language = language
+
+
 class WhisperX:
     def __init__(self, config: WhisperXPipelineConfig):
         self.config = config
 
-    def __call__(self, audio_path: Path | str) -> pd.DataFrame:
+    def __call__(self, input: WhisperXInput) -> pd.DataFrame:
+        audio_path = input.audio_path
         if isinstance(audio_path, str):
             audio_path = Path(audio_path)
 
@@ -72,6 +81,10 @@ def __call__(self, audio_path: Path | str) -> pd.DataFrame:
             "--diarize",
         ]
 
+        # Add language if provided
+        if input.language:
+            args.extend(["--language", input.language])
+
         # Run whisperx CLI
         subprocess.run(args)
 
@@ -127,11 +140,19 @@ class WhisperXPipeline(Pipeline):
     _config_class = WhisperXPipelineConfig
     pipeline_type = PipelineType.ORCHESTRATION
 
-    def build_pipeline(self) -> Callable[[Path], pd.DataFrame]:
+    def build_pipeline(self) -> Callable[[WhisperXInput], pd.DataFrame]:
         return WhisperX(self.config)
 
-    def parse_input(self, input_sample: DiarizationSample) -> Path:
-        return input_sample.save_audio(TEMP_AUDIO_DIR)
+    def parse_input(self, input_sample: DiarizationSample) -> WhisperXInput:
+        # Extract language if force_language is enabled
+        language = None
+        if self.config.force_language:
+            language = input_sample.language
+
+        return WhisperXInput(
+            audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
+            language=language,
+        )
 
     def parse_output(self, output: pd.DataFrame) -> OrchestrationOutput:
         output = output.assign(words=lambda df: df["text"].str.split()).explode("words")
diff --git a/src/openbench/pipeline/transcription/apple_speech_analyzer.py b/src/openbench/pipeline/transcription/apple_speech_analyzer.py
index 7153cc7..992658b 100644
--- a/src/openbench/pipeline/transcription/apple_speech_analyzer.py
+++ b/src/openbench/pipeline/transcription/apple_speech_analyzer.py
@@ -122,9 +122,15 @@ def build_pipeline(self) -> Callable[[SpeechAnalyzerCliInput], Path]:
         return engine.transcribe
 
     def parse_input(self, input_sample: TranscriptionSample) -> SpeechAnalyzerCliInput:
+        # Extract language if force_language is enabled
+        language = None
+        if self.config.force_language:
+            language = input_sample.language
+
         return SpeechAnalyzerCliInput(
             audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
             keep_audio=False,
+            language=language,
         )
 
     def parse_output(self, output: Path) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/transcription_assemblyai.py b/src/openbench/pipeline/transcription/transcription_assemblyai.py
index 9c7a07b..bed1612 100644
--- a/src/openbench/pipeline/transcription/transcription_assemblyai.py
+++ b/src/openbench/pipeline/transcription/transcription_assemblyai.py
@@ -99,6 +99,13 @@ def parse_input(self, input_sample: TranscriptionSample) -> Path:
             if keywords:
                 self.current_keywords = keywords
 
+        # Warn if force_language is enabled (not currently supported)
+        if self.config.force_language:
+            logger.warning(
+                f"{self.__class__.__name__} does not support language hinting. "
+                "The force_language flag will be ignored."
+            )
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: str) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/transcription_deepgram.py b/src/openbench/pipeline/transcription/transcription_deepgram.py
index 02dc146..75973fa 100644
--- a/src/openbench/pipeline/transcription/transcription_deepgram.py
+++ b/src/openbench/pipeline/transcription/transcription_deepgram.py
@@ -27,10 +27,19 @@ class DeepgramTranscriptionPipeline(Pipeline):
     pipeline_type = PipelineType.TRANSCRIPTION
 
     def build_pipeline(self) -> Callable[[Path], DeepgramApiResponse]:
-        deepgram_api = DeepgramApi(options=PrerecordedOptions(model=self.config.model_version, smart_format=True))
+        # Create base API without language detection
+        base_api = DeepgramApi(
+            options=PrerecordedOptions(
+                model=self.config.model_version, smart_format=True, detect_language=not self.config.force_language
+            )
+        )
 
         def transcribe(audio_path: Path) -> DeepgramApiResponse:
-            response = deepgram_api.transcribe(audio_path, keyterm=self.current_keywords)
+            # Use language-specific API if language is set, otherwise use base API
+            if self.current_language:
+                base_api.set_language(self.current_language)
+
+            response = base_api.transcribe(audio_path, keyterm=self.current_keywords)
             # Remove temporary audio path
             audio_path.unlink(missing_ok=True)
             return response
@@ -38,7 +47,7 @@ def transcribe(audio_path: Path) -> DeepgramApiResponse:
         return transcribe
 
     def parse_input(self, input_sample) -> Path:
-        """Override to extract keywords from sample before processing."""
+        """Override to extract keywords and language from sample before processing."""
         self.current_keywords = None
         if self.config.use_keywords:
             keywords = input_sample.extra_info.get("dictionary", [])
@@ -46,6 +55,11 @@ def parse_input(self, input_sample) -> Path:
                 # Add + between keywords for Deepgram URL
                 self.current_keywords = "+".join(keywords)
 
+        # Extract language if force_language is enabled
+        self.current_language = None
+        if self.config.force_language:
+            self.current_language = input_sample.language
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: DeepgramApiResponse) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/transcription_nemo.py b/src/openbench/pipeline/transcription/transcription_nemo.py
index 7556946..68a1205 100644
--- a/src/openbench/pipeline/transcription/transcription_nemo.py
+++ b/src/openbench/pipeline/transcription/transcription_nemo.py
@@ -219,6 +219,13 @@ def parse_input(self, input_sample) -> Path:
                     self.context_graph = context_biasing.ContextGraphCTC(blank_id=self.blank_idx)
                     self.context_graph.add_to_graph(context_transcripts)
 
+        # Warn if force_language is enabled (not currently supported)
+        if self.config.force_language:
+            logger.warning(
+                f"{self.__class__.__name__} does not support language hinting. "
+                "The force_language flag will be ignored."
+            )
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: TranscriptionOutput) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/transcription_openai.py b/src/openbench/pipeline/transcription/transcription_openai.py
index f12de12..1c6d6aa 100644
--- a/src/openbench/pipeline/transcription/transcription_openai.py
+++ b/src/openbench/pipeline/transcription/transcription_openai.py
@@ -33,6 +33,7 @@ def transcribe(audio_path: Path) -> TranscriptionVerbose:
             response = openai_api.transcribe(
                 audio_path,
                 prompt=self.current_keywords_prompt,
+                language=self.current_language,
             )
             # Remove temporary audio path
             audio_path.unlink(missing_ok=True)
@@ -41,7 +42,7 @@ def transcribe(audio_path: Path) -> TranscriptionVerbose:
         return transcribe
 
     def parse_input(self, input_sample) -> Path:
-        """Override to extract keywords from sample before processing."""
+        """Override to extract keywords and language from sample before processing."""
         # Extract keywords from sample's extra_info if flag is enabled
         self.current_keywords_prompt = None
         if self.config.use_keywords:
@@ -50,6 +51,11 @@ def parse_input(self, input_sample) -> Path:
                 # Format keywords as comma-separated prompt for OpenAI
                 self.current_keywords_prompt = ", ".join(keywords)
 
+        # Extract language if force_language is enabled
+        self.current_language = None
+        if self.config.force_language:
+            self.current_language = input_sample.language
+
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
     def parse_output(self, output: TranscriptionVerbose) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/transcription_oss_whisper.py b/src/openbench/pipeline/transcription/transcription_oss_whisper.py
index 8bda785..622f549 100644
--- a/src/openbench/pipeline/transcription/transcription_oss_whisper.py
+++ b/src/openbench/pipeline/transcription/transcription_oss_whisper.py
@@ -157,7 +157,7 @@ def parse_input(self, input_sample: TranscriptionSample) -> Path:
         # Extract language if force_language is enabled
         self.current_language = None
         if self.config.force_language:
-            self.current_language = input_sample.extra_info.get("language", None)
+            self.current_language = input_sample.language
 
         return input_sample.save_audio(TEMP_AUDIO_DIR)
 
diff --git a/src/openbench/pipeline/transcription/transcription_whisperkitpro.py b/src/openbench/pipeline/transcription/transcription_whisperkitpro.py
index f57213e..bc0f88a 100644
--- a/src/openbench/pipeline/transcription/transcription_whisperkitpro.py
+++ b/src/openbench/pipeline/transcription/transcription_whisperkitpro.py
@@ -110,7 +110,7 @@ def build_pipeline(self) -> WhisperKitPro:
         return engine
 
     def parse_input(self, input_sample: TranscriptionSample) -> WhisperKitProInput:
-        """Override to extract keywords from sample before processing."""
+        """Override to extract keywords and language from sample before processing."""
         # Extract keywords from sample's extra_info if flag is enabled
         custom_vocab_path = None
         if self.config.use_keywords:
@@ -129,10 +129,16 @@ def parse_input(self, input_sample: TranscriptionSample) -> WhisperKitProInput:
                 custom_vocab_path = str(vocab_file)
                 logger.debug(f"Created custom vocabulary file: {custom_vocab_path} with {len(keywords)} keywords")
 
+        # Extract language if force_language is enabled
+        language = None
+        if self.config.force_language:
+            language = input_sample.language
+
         return WhisperKitProInput(
             audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
             keep_audio=False,
             custom_vocabulary_path=custom_vocab_path,
+            language=language,
         )
 
     def parse_output(self, output: WhisperKitProOutput) -> TranscriptionOutput:
diff --git a/src/openbench/pipeline/transcription/whisperkit.py b/src/openbench/pipeline/transcription/whisperkit.py
index 789ab7d..2e0f444 100644
--- a/src/openbench/pipeline/transcription/whisperkit.py
+++ b/src/openbench/pipeline/transcription/whisperkit.py
@@ -257,9 +257,15 @@ def build_pipeline(self) -> Callable[[TranscriptionCliInput], TranscriptionCliOu
         return engine.transcribe
 
     def parse_input(self, input_sample: TranscriptionSample) -> TranscriptionCliInput:
+        # Extract language if force_language is enabled
+        language = None
+        if self.config.force_language:
+            language = input_sample.language
+
         return TranscriptionCliInput(
             audio_path=input_sample.save_audio(TEMP_AUDIO_DIR),
             keep_audio=False,
+            language=language,
         )
 
     def parse_output(self, output: TranscriptionCliOutput) -> TranscriptionOutput: