Chirp 2: Enhanced multilingual accuracy

Chirp 2 is the latest generation of Google's multilingual ASR-specific models, designed to meet user needs based on feedback and experience. It improves on the original Chirp model in accuracy and speed, as well as expanding into key new features like word-level timestamps, model adaptation, and speech translation.

Model details

Chirp 2 is exclusively available within the Speech-to-Text API V2.

Model identifiers

You can utilize Chirp 2 just like any other model by specifying the appropriate model identifier in your recognition request when using the API or the model name while in the Google Cloud console.

Model	Model identifier
Chirp 2	chirp_2

API methods

As Chirp 2 is exclusively available within the Speech-to-Text API V2, it supports the following recognition methods:

Model	Model identifier	Language support
V2	Speech.StreamingRecognize (good for streaming and real-time audio)	Limited*
V2	Speech.Recognize (good for short audio < 1 min)	On par with Chirp
V2	Speech.BatchRecognize (good for long audio 1 min to 8 hrs)	On par with Chirp

*You can always find the latest list of supported languages and features for each transcription model, using the locations API.

Regional availability

Chirp 2 is supported in the following regions:

Google Cloud Zone	Launch Readiness
us-central1	GA
europe-west4	GA
asia-southeast1	Private Preview

You can always find the latest list of supported Google Cloud regions, languages, and features for each transcription model, using the locations API as explained here.

Language availability for transcription

Refer to our list of supported transcription languages.

Language availability for translation

These are the supported languages for speech translation. Note that Chirp 2's language support for translation is not symmetrical. This means that while we may be able to translate from Language A to Language B, translation from Language B to Language A might not be available. The following language pairs are supported for Speech-Translation.

For translation to English:

Source -> Target Language	Source -> Target Language Code
Arabic (Egypt) -> English	ar-EG -> en-US
Arabic (Gulf) -> English	ar-x-gulf -> en-US
Arabic (Levant) -> English	ar-x-levant -> en-US
Arabic (Maghrebi) -> English	ar-x-maghrebi -> en-US
Catalan (Spain) -> English	ca-ES -> en-US
Welsh (United Kingdom) -> English	cy-GB -> en-US
German (Germany) -> English	de-DE -> en-US
Spanish (Latin America) -> English	es-419 -> en-US
Spanish (Spain) -> English	es-ES -> en-US
Spanish (United States) -> English	es-US -> en-US
Estonian (Estonia) -> English	et-EE -> en-US
French (Canada) -> English	fr-CA -> en-US
French (France) -> English	fr-FR -> en-US
Persian (Iran) -> English	fa-IR -> en-US
Indonesian (Indonesia) -> English	id-ID -> en-US
Italian (Italy) -> English	it-IT -> en-US
Japanese (Japan) -> English	ja-JP -> en-US
Latvian (Latvia) -> English	lv-LV -> en-US
Mongolian (Mongolia) -> English	mn-MN -> en-US
Dutch (Netherlands) -> English	nl-NL -> en-US
Portuguese (Brazil) -> English	pt-BR -> en-US
Russian (Russia) -> English	ru-RU -> en-US
Slovenian (Slovenia) -> English	sl-SI -> en-US
Swedish (Sweden) -> English	sv-SE -> en-US
Tamil (India) -> English	ta-IN -> en-US
Turkish (Turkey) -> English	tr-TR -> en-US
Chinese (Simplified, China) -> English	zh-Hans-CN -> en-US

For translation from English:

Source -> Target Language	Source -> Target Language Code
English -> Arabic (Egypt)	en-US -> ar-EG
English -> Arabic (Gulf)	en-US -> ar-x-gulf
English -> Arabic (Levant)	en-US -> ar-x-levant
English -> Arabic (Maghrebi)	en-US -> ar-x-maghrebi
English -> Catalan (Spain)	en-US -> ca-ES
English -> Welsh (United Kingdom)	en-US -> cy-GB
English -> German (Germany)	en-US -> de-DE
English -> Estonian (Estonia)	en-US -> et-EE
English -> Persian (Iran)	en-US -> fa-IR
English -> Indonesian (Indonesia)	en-US -> id-ID
English -> Japanese (Japan)	en-US -> ja-JP
English -> Latvian (Latvia)	en-US -> lv-LV
English -> Mongolian (Mongolia)	en-US -> mn-MN
English -> Slovenian (Slovenia)	en-US -> sl-SI
English -> Swedish (Sweden)	en-US -> sv-SE
English -> Tamil (India)	en-US -> ta-IN
English -> Turkish (Turkey)	en-US -> tr-TR
English -> Chinese (Simplified, China)	en-US -> zh-Hans-CN

Feature support and limitations

Chirp 2 supports the following features:

Feature	Description
Automatic punctuation	Automatically generated by the model and can be optionally disabled.
Automatic capitalization	Automatically generated by the model and can be optionally disabled.
Speech adaptation (Biasing)	Provide hints to the model in the form of simple words or phrases to improve recognition accuracy for specific terms or proper nouns. Class tokens or custom classes are not supported.
Word-timings (Timestamps)	Automatically generated by the model and can be optionally enabled. It is possible that transcrption quality and speed will degrade slightly.
Language-agnostic audio transcription	The model automatically infers the spoken language in your audio file and transcribes in the most prevalent language.
Language-specific translation	The model automatically translates from the spoken language to the target language.
Forced normalization	If defined in the request body, the API will perform string replacements on specific terms or phrases, ensuring consistency in the transcription.
Word-level confidence scores	The API returns a value, but it isn't truly a confidence score. In the case of translation, confidence scores are not returned.

Chirp 2 does not support the following features:

Feature	Description
Diarization	Not supported
Language Detection	Not supported
Profanity filter	Not supported

Transcribe using Chirp 2

Discover how to use Chirp 2 for your transcription and translation needs.

Perform streaming speech recognition

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_streaming_chirp2(
    audio_file: str
) -> cloud_speech.StreamingRecognizeResponse:
    """Transcribes audio from audio file stream using the Chirp 2 model of Google Cloud Speech-to-Text V2 API.

    Args:
        audio_file (str): Path to the local audio file to be transcribed.
            Example: "resources/audio.wav"

    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API V2 containing
        the transcription results.        
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        content = f.read()

    # In practice, stream should be a generator yielding chunks of audio data
    chunk_length = len(content) // 5
    stream = [
        content[start : start + chunk_length]
        for start in range(0, len(content), chunk_length)
    ]
    audio_requests = (
        cloud_speech.StreamingRecognizeRequest(audio=audio) for audio in stream
    )

    recognition_config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),        
        language_codes=["en-US"],
        model="chirp_2",
    )
    streaming_config = cloud_speech.StreamingRecognitionConfig(
        config=recognition_config
    )
    config_request = cloud_speech.StreamingRecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        streaming_config=streaming_config,
    )

    def requests(config: cloud_speech.RecognitionConfig, audio: list) -> list:
        yield config
        yield from audio

    # Transcribes the audio into text
    responses_iterator = client.streaming_recognize(
        requests=requests(config_request, audio_requests)
    )
    responses = []
    for response in responses_iterator:
        responses.append(response)
        for result in response.results:
            print(f"Transcript: {result.alternatives[0].transcript}")

    return responses

Perform synchronous speech recognition

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_sync_chirp2(
    audio_file: str
) -> cloud_speech.RecognizeResponse:
    """Transcribes an audio file using the Chirp 2 model of Google Cloud Speech-to-Text V2 API.
    Args:
        audio_file (str): Path to the local audio file to be transcribed.
            Example: "resources/audio.wav"
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the transcription results.
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        audio_content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["en-US"],
        model="chirp_2",
    )

    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        content=audio_content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Transcript: {result.alternatives[0].transcript}")

    return response

Perform batch speech recognition

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_batch_chirp2(
    audio_uri: str,
) -> cloud_speech.BatchRecognizeResults:
    """Transcribes an audio file from a Google Cloud Storage URI using the Chirp 2 model of Google Cloud Speech-to-Text V2 API.
    Args:
        audio_uri (str): The Google Cloud Storage URI of the input audio file.
            E.g., gs://[BUCKET]/[FILE]
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the transcription results.
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["en-US"],
        model="chirp_2",
    )

    file_metadata = cloud_speech.BatchRecognizeFileMetadata(uri=audio_uri)

    request = cloud_speech.BatchRecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        files=[file_metadata],
        recognition_output_config=cloud_speech.RecognitionOutputConfig(
            inline_response_config=cloud_speech.InlineOutputConfig(),
        ),
    )

    # Transcribes the audio into text    
    operation = client.batch_recognize(request=request)

    print("Waiting for operation to complete...")
    response = operation.result(timeout=120)

    for result in response.results[audio_uri].transcript.results:
        print(f"Transcript: {result.alternatives[0].transcript}")

    return response.results[audio_uri].transcript

Use Chirp 2 Features

Explore how you can use the latest features, with code examples:

Perform a language-agnostic transcription

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_sync_chirp2_auto_detect_language(
    audio_file: str
) -> cloud_speech.RecognizeResponse:
    """Transcribes an audio file and auto-detect spoken language using Chirp 2.
    Please see https://cloud.google.com/speech-to-text/v2/docs/encoding for more
    information on which audio encodings are supported.
    Args:
        audio_file (str): Path to the local audio file to be transcribed.
            Example: "resources/audio.wav"
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the transcription results.
    """
    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        audio_content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["auto"],  # Set language code to auto to detect language.
        model="chirp_2",
    )

    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        content=audio_content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Transcript: {result.alternatives[0].transcript}")
        print(f"Detected Language: {result.language_code}")

    return response

Perform speech translation

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def translate_sync_chirp2(
    audio_file: str
) -> cloud_speech.RecognizeResponse:
    """Translates an audio file using Chirp 2.
    Args:
        audio_file (str): Path to the local audio file to be translated.
            Example: "resources/audio.wav"
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the translated results.
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        audio_content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["fr-FR"],  # Set language code to targeted to detect language.
        translation_config=cloud_speech.TranslationConfig(target_language="fr-FR"), # Set target language code.
        model="chirp_2",
    )

    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        content=audio_content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Translated transcript: {result.alternatives[0].transcript}")

    return response

Enable word-level timestamps

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_sync_chirp2_with_timestamps(
    audio_file: str
) -> cloud_speech.RecognizeResponse:
    """Transcribes an audio file using the Chirp 2 model of Google Cloud Speech-to-Text V2 API, providing word-level timestamps for each transcribed word. 
    Args:
        audio_file (str): Path to the local audio file to be transcribed.
            Example: "resources/audio.wav"
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the transcription results.
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        audio_content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["en-US"],
        model="telephony",
        features=cloud_speech.RecognitionFeatures(
            enable_word_time_offsets=True, # Enabling word-level timestamps
        )
    )

    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        content=audio_content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Transcript: {result.alternatives[0].transcript}")

    return response

Improve accuracy with model adaptation

import os

from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech
from google.api_core.client_options import ClientOptions

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT")

def transcribe_sync_chirp2_model_adaptation(
    audio_file: str
) -> cloud_speech.RecognizeResponse:
    """Transcribes an audio file using the Chirp 2 model with adaptation, improving accuracy for specific audio characteristics or vocabulary.
    Args:
        audio_file (str): Path to the local audio file to be transcribed.
            Example: "resources/audio.wav"
    Returns:
        cloud_speech.RecognizeResponse: The response from the Speech-to-Text API containing
        the transcription results.
    """

    # Instantiates a client
    client = SpeechClient(
        client_options=ClientOptions(
            api_endpoint="us-central1-speech.googleapis.com",
        )
    )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        audio_content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["en-US"],
        model="chirp_2",
        # Use model adaptation
        adaptation=cloud_speech.SpeechAdaptation( 
          phrase_sets=[
              cloud_speech.SpeechAdaptation.AdaptationPhraseSet(
                  inline_phrase_set=cloud_speech.PhraseSet(phrases=[
                    {
                        "value": "alphabet",
                    },
                    {
                          "value": "cell phone service",
                    }
                  ])
              )
          ]
        )
    )

    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/{PROJECT_ID}/locations/us-central1/recognizers/_",
        config=config,
        content=audio_content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Transcript: {result.alternatives[0].transcript}")

    return response

Use Chirp 2 in the Google Cloud console

1. Ensure you have signed up for a Google Cloud account and created a project.
2. Go to Speech in Google Cloud console.
3. Enable the API if it's not already enabled.

4. Make sure that you have an STT console Workspace. If you do not have one already, you need to create a workspace.

   1. Visit the transcriptions page, and click New Transcription.

   2. Open the Workspace drop-down and click New Workspace to create a workspace for transcription.

   3. From the Create a new workspace navigation sidebar, click Browse.

   4. Click to create a new bucket.

   5. Enter a name for your bucket and click Continue.

   6. Click Create to create your Cloud Storage bucket.

   7. After the bucket is created, click Select to select your bucket for use.

   8. Click Create to finish creating your workspace for the Speech-to-Text API V2 console.

5. Perform a transcription on your actual audio.

   

   From the New Transcription page, select your audio file through either upload (Local upload) or specifying an existing Cloud Storage file (Cloud storage).

6. Click Continue to move to the Transcription options.

   

   1. Select the Spoken language that you plan to use for recognition with Chirp from your previously created recognizer.

   2. In the model drop-down, select Chirp - Universal Speech Model.

   3. In the Recognizer drop-down, select your newly created recognizer.

   4. Click Submit to run your first recognition request using Chirp.

7. View your Chirp 2 transcription result.

   1. From the Transcriptions page, click the name of the transcription to view its result.

   2. In the Transcription details page, view your transcription result, and optionally playback the audio in the browser.

Clean up

To avoid incurring charges to your Google Cloud account for the resources used on this page, follow these steps.

1. Optional: Revoke the authentication credentials that you created, and delete the local credential file.

   gcloud auth application-default revoke

2. Optional: Revoke credentials from the gcloud CLI.

   gcloud auth revoke

Console

Caution: Deleting a project has the following effects:

• Everything in the project is deleted. If you used an existing project for the tasks in this document, when you delete it, you also delete any other work you've done in the project.
• Custom project IDs are lost. When you created this project, you might have created a custom project ID that you want to use in the future. To preserve the URLs that use the project ID, such as an appspot.com URL, delete selected resources inside the project instead of deleting the whole project.

If you plan to explore multiple architectures, tutorials, or quickstarts, reusing projects can help you avoid exceeding project quota limits.

In the Google Cloud console, go to the Manage resources page.

Go to Manage resources

In the project list, select the project that you want to delete, and then click Delete.

In the dialog, type the project ID, and then click Shut down to delete the project.

gcloud

Caution: Deleting a project has the following effects:

• Everything in the project is deleted. If you used an existing project for the tasks in this document, when you delete it, you also delete any other work you've done in the project.
• Custom project IDs are lost. When you created this project, you might have created a custom project ID that you want to use in the future. To preserve the URLs that use the project ID, such as an appspot.com URL, delete selected resources inside the project instead of deleting the whole project.

If you plan to explore multiple architectures, tutorials, or quickstarts, reusing projects can help you avoid exceeding project quota limits.

Delete a Google Cloud project:

gcloud projects delete PROJECT_ID

What's next

• Practice transcribing short audio files.
• Learn how to transcribe streaming audio.
• Learn how to transcribe long audio files.
• For best performance, accuracy, and other tips, see the best practices documentation.