Google Speech-to-Text component

Q: How do I transcribe a file stored in Cloud Storage?

Send its URI as the body text, for example gs://my-bucket/call.wav.

Q: How do I transcribe long recordings?

Use operation=longRunningRecognize, then poll with operation=getOperation (for example from a timer route) until CamelGoogleSpeechToTextDone is true.

Q: Is streaming recognition supported?

No. Only the recognize and longrunningrecognize REST calls are implemented.

The google-speech-to-text component transcribes audio to text with the Google Cloud Speech-to-Text v1 API, from audio bytes or a gs:// file.

Scheme: google-speech-to-text: · Crate: cameleer-cloud · Cargo feature: cloud · Consumer: no · Producer: yes · Apache Camel: none (Cameleer component)

When to use it

Transcribe voice notes, call recordings or voicemail dropped into a folder or bucket.
Turn spoken commands from an edge device into text, then route on the text.
Feed transcripts to an LLM for summarization or classification (OpenAI, Vertex AI).

URI format

text

google-speech-to-text://label[?options]

The label is free text that names the endpoint; it is not sent to the API.

Example

Transcribe every .wav file in a folder and write the transcript next to it as a .txt file:

yaml example: google-speech-to-text-transcribe-folder

- route:
    id: transcribe-voice-notes
    from:
      uri: "file:{{env:INBOX_DIR:voice-notes}}?include=.*\\.wav&move=.done"
      steps:
        - to: "google-speech-to-text://notes?languageCode={{env:LANGUAGE:en-US}}&sampleRateHertz=16000&enableAutomaticPunctuation=true"
        - log: "transcribed ${file:name} (confidence ${header.CamelGoogleSpeechToTextConfidence})"
        - setHeader:
            name: CamelFileName
            simple: "${file:onlyname.noext}.txt"
        - to: "file:{{env:OUT_DIR:transcripts}}"

Rust DSL

rust

ctx.add_routes(|b| {
    b.from("file:voice-notes?include=.*\\.wav")
        .to("google-speech-to-text://notes?sampleRateHertz=16000")
        .log("${body}");
})?;

Options

option	default	description
`operation`	`recognize`	`recognize`, `longRunningRecognize`, `getOperation`. Header `CamelGoogleSpeechToTextOperation`.
`encoding`	`LINEAR16`	Audio encoding (`LINEAR16`, `FLAC`, `MULAW`, `OGG_OPUS`, ...).
`sampleRateHertz`		Sample rate of the audio.
`languageCode`	`en-US`	Language of the speech.
`model`		Recognition model, for example `latest_long` or `phone_call`.
`enableAutomaticPunctuation`	`false`	Add punctuation to the transcript.

Credentials are shared by all Google components: see credentials.

Operations

recognize: the body is the audio bytes, or a text gs://bucket/file URI. Body = the best transcript (results joined with spaces).
longRunningRecognize: starts an asynchronous recognition (for audio longer than a minute); body and CamelGoogleSpeechToTextOperationName = the operation name.
getOperation: checks the operation in CamelGoogleSpeechToTextOperationName; sets CamelGoogleSpeechToTextDone, and the body is the transcript once done.

Message headers

header	direction	description
`CamelGoogleSpeechToTextOperation`	in	Operation.
`CamelGoogleSpeechToTextConfidence`	out	Confidence of the first result.
`CamelGoogleSpeechToTextResults`	out	All results returned by the API.
`CamelGoogleSpeechToTextOperationName`	in / out	Long-running operation name.
`CamelGoogleSpeechToTextDone`	out	`getOperation`: whether the operation finished.

Run it with Docker

docker run --rm -v "$PWD:/routes:ro" -v "$PWD/voice-notes:/in" -v "$PWD/transcripts:/out" \
  -v "$HOME/keys/sa.json:/keys/sa.json:ro" -e GOOGLE_APPLICATION_CREDENTIALS=/keys/sa.json \
  -e INBOX_DIR=/in -e OUT_DIR=/out \
  pmsudhi/cameleer run /routes/transcribe-voice-notes.camel.yaml

Common questions

How do I transcribe a file stored in Cloud Storage?

Send its URI as the body text, for example gs://my-bucket/call.wav.

How do I transcribe long recordings?

Use operation=longRunningRecognize, then poll with operation=getOperation (for example from a timer route) until CamelGoogleSpeechToTextDone is true.

Is streaming recognition supported?

No. Only the recognize and longrunningrecognize REST calls are implemented.

Found a problem?

If something on this page doesn't match what Cameleer does, or you hit a bug while building with it, please help fix it, whether you are a person or a coding agent: open an issue that describes what you saw, then fork the repository, fix it, add or update a test, run cargo test --workspace, and open a pull request. CONTRIBUTING.md explains the ground rules.