enhanced_speech_to_text_built_on_whisper.proto

path phonexia/grpc/technologies/enhanced_speech_to_text_built_on_whisper/v1/enhanced_speech_to_text_built_on_whisper.proto

package phonexia.grpc.technologies.enhanced_speech_to_text_built_on_whisper.v1

Phonexia Enhanced Speech To Text Built On Whisper gRPC API.

---

Messages

ListSupportedLanguagesRequest

The top-level message sent by the client for the ListSupportedLanguages method.

Name	Type	Description

ListSupportedLanguagesResponse

The top-level message returned to the client by the ListSupportedLanguages

Name	Type	Description
languages	repeated string	List of available languages for transcription. The language is specified as an RFC 5646 language tag.
translation_languages	repeated string	List of available source languages for translation. The language is specified as an RFC 5646 language tag.

OneBest

The one-best output from Speech To Text.

Name	Type	Description
segments	repeated OneBestSegment	Sequential list of transcription segments corresponding to sequential portions of audio.

OneBestSegment

Segment of the one-best output.

Name	Type	Description
text	string	Text content of the segment.
start_time	google.protobuf.Duration	Start time of the segment.
end_time	google.protobuf.Duration	End time of the segment.
language	string	Target language, that the text field of this message is transcribed in. The language is specified as an RFC 5646 language tag. Note: If a specific language was enforced in the Transcribe method, the target language is the same as this specified language.
source_language	string	The selected language of the audio data as an RFC 5646 language tag. This field contains the language that whisper used for transcription/translation. In case that the user does not have a license for the language identified in the audio data, this field may not contain this language. Rather it will contain the most probable of the licensed languages. Note that this will have radical influence on the output, so always compare this field against detected_source_language to know if the text is transcribed/translated to/from the correct language. Note: If a specific source_language was enforced in the config, this field is the same as that specified language.
detected_source_language	string	The detected language of the audio data as an RFC 5646 language tag. This field contains the language that whisper detected as the source language of the audio data. This field may not equal to source_language in case the audio data is in different language than from the set of licensed languages. Always check if this field is equal to source_language.
words	repeated Word	Detailed word-level segmentation of the segment.

TranscribeConfig

Name	Type	Description
language	string	The language of the audio data specified as an RFC 5646 language tag. If not set, the language will be detected automatically.
enable_language_switching	bool	By default, the language of the speech is detected once at the beginning of the audio. Setting enable_language_switching to true allows for dynamic language switching in the audio, with the language being detected approximately every 30 seconds. Note: This option is ignored if the language is enforced by setting the language option.
enable_word_segmentation	bool	If set to true, the result will contain word-level segmentation. Note: Enabling this option may incrase processing time.

TranscribeRequest

The top-level message sent by the client for the Transcribe method.

Name	Type	Description
audio	phonexia.grpc.common.Audio	Audio data from which a transcription should be extracted. A variety of common audio formats are supported, including WAV, FLAC, and others. There is no minimum audio length limit.
config	TranscribeConfig	Speech to text transcription configuration.

TranscribeResponse

The top-level message returned to the client by the Transcribe method.

Name	Type	Description
result	TranscribeResult	The Speech to Text transcription result.
processed_audio_length	google.protobuf.Duration	Total length of the processed audio. Set only if this is the last response in the stream.

TranscribeResult

The collection of all transcription result formats.

Name	Type	Description
one_best	OneBest	Result in one-best format.

TranslateConfig

Name	Type	Description
source_language	string	The language of the audio data specified as an RFC 5646 language tag. If not set, the language will be detected automatically.
enable_language_switching	bool	By default, the language of the speech is detected once at the beginning of the audio. Setting enable_language_switching to true allows for dynamic language switching in the audio, with the language being detected approximately every 30 seconds. Note: This option is ignored if the language is enforced by setting the source_language option.
enable_word_segmentation	bool	If set to true, the result will contain word-level segmentation. Note: Enabling this option may increase processing time.

TranslateRequest

The top-level message sent by the client for the Translate method.

Name	Type	Description
audio	phonexia.grpc.common.Audio	Audio data on which the translation should be performed. A variety of common audio formats are supported, including WAV, FLAC, and others. There is no minimum audio length limit.
config	TranslateConfig	Speech to text translation configuration.

TranslateResponse

The top-level message returned to the client by the Translate method.

Name	Type	Description
result	TranslateResult	The Speech to Text translation result.
processed_audio_length	google.protobuf.Duration	Total length of the processed audio. Set only if this is the last response in the stream.

TranslateResult

The collection of all translation result formats.

Name	Type	Description
one_best	OneBest	Result in one-best format.

Word

Individual word with timestamps.

Name	Type	Description
text	string	Text of the word.
start_time	google.protobuf.Duration	Start time of the word.
end_time	google.protobuf.Duration	End time of the word.
punctuated_text	string	Text of the word with punctuation and capitalization.

---

Services

SpeechToText

Service implementing speech-to-text transcription.

Transcribe

Method	Transcribe
Request	TranscribeRequest stream
Response	TranscribeResponse stream
Description	Performs synchronous speech-to-text transcription. The transcription starts after the whole audio has been received.

Translate

Method	Translate
Request	TranslateRequest stream
Response	TranslateResponse stream
Description	Performs synchronous speech-to-text translation. The translation starts after the whole audio has been received. Currently, the target language is limited to English only. The set of available source languages can be found by invoking the ListSupportedLanguages method.

ListSupportedLanguages

Method	ListSupportedLanguages
Request	ListSupportedLanguagesRequest
Response	ListSupportedLanguagesResponse
Description	Method to retrieve supported languages for transcription/translation. Models can be monolingual or multilingual. Multilingual models may be licensed to only a subset of all supported languages, in which case this method returns this limited subset.

---