Skip to main content

Client for alvenir gRPC streaming services

Project description

Alvenir Client for Speech Recognition (BETA)

gRPC client for speech recognition using Alvenir as backend.

The client and the gRPC integration is in beta and ongoing development. Hence, there might be a few stability issues. It currently supports only Danish and you need to receive an API key from Alvenir to use.

Write an email to martin[at]alvenir.ai or rasmus[at]alvenir.ai if you want an API key!

How to install

Installing the client is as easy as writing:

pip install alvenirclient

How to use

It is very simple to use! The preferred format of audio is single channel 16kHz wav with pcm_s16le audio encoding.

If you audio files is in different format, you can use a tool such as ffmpeg to do the transcoding. An example is:

ffmpeg -i <existing_audio_file_path> -acodec pcm_s16le -ar 16000 -ac 1 <new_wav_audio_file_path>

The server can also perform transcoding but this feature is still experimential.

Basic usage

from alvenirclient.client import AlvenirClient

client = AlvenirClient(key="<YOUR KEY HERE>")
response = client.transcribe_file("<Path to file>", server_transcoding=False)
print(response.transcription)
print(response.full_document)

The response is of class AlvenirResponse and contains:

  • transcription; The transcription as a string
  • full_document; A json document with confidences and timestamps. An example of a full document is
{
  "confidence": 0.9141539366420111,
  "segments": [
    {
      "confidence": 0.9141539366420111,
      "end_time": 1.77,
      "start_time": 0.93,
      "word_list": [
        {
          "confidence": 0.995308015467965,
          "end_time": 1.01,
          "start_time": 0.93,
          "word": "Jeg"
        },
        {
          "confidence": 0.9983347185448589,
          "end_time": 1.1500000000000001,
          "start_time": 1.03,
          "word": "bor"
        },
        {
          "confidence": 0.9622011164767056,
          "end_time": 1.29,
          "start_time": 1.21,
          "word": "hos"
        },
        {
          "confidence": 0.9999113092492331,
          "end_time": 1.4500000000000002,
          "start_time": 1.37,
          "word": "min"
        },
        {
          "confidence": 0.6150145234712928,
          "end_time": 1.77,
          "start_time": 1.51,
          "word": "datter."
        }
      ]
    }
  ]
}

Server transcoding

Performing server_transcoding just requires setting server_transcoding=True. This way, the samplerate and fileformat does not matter as long as it is a common audio format (wav, mp3 etc.).

from alvenirclient.client import AlvenirClient

client = AlvenirClient(key="<YOUR KEY HERE>")
response = client.transcribe_file("<Path to file>", server_transcoding=True)
print(response.transcription)
print(response.full_document)

Microphone transcription

The client allows you to use your own microphone for transcriptions. It does however require you to additionally install pyaudio. It is a very experimential feature and the transcriptions might not be very accurate depending on the microphone and how the audio is recorded.

The most simple usage with microphone is:

from alvenirclient.client import AlvenirClient

client = AlvenirClient(key="<YOUR KEY HERE>")
response = client.transcribe_microphone()
print(response.transcription)
print(response.full_document)

The microphone will stop recording when you press the key "enter". This will not be real time, but the transcription is being made while you speak, so the response should be pretty instant.

For real time transcription i.e. continously getting a response, use the following code.

from alvenirclient.client import AlvenirClient
from alvenirclient.audio_pb2 import STOP

client = AlvenirClient(key="<YOUR KEY HERE>")
final_response = None
responses_iterable = client.transcribe_microphone_realtime()

for response in responses_iterable:
    if response.status == STOP:
        final_response = response
        break
    else:
        print(response.transcription)

print(final_response.transcription)
print(final_response.full_document)

Note, the intermediate responses do not have full_document but only text in the transcription.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

alvenirclient-0.0.2.tar.gz (11.6 kB view details)

Uploaded Source

File details

Details for the file alvenirclient-0.0.2.tar.gz.

File metadata

  • Download URL: alvenirclient-0.0.2.tar.gz
  • Upload date:
  • Size: 11.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.8.0 pkginfo/1.8.2 readme-renderer/34.0 requests/2.27.1 requests-toolbelt/0.9.1 urllib3/1.26.8 tqdm/4.62.3 importlib-metadata/3.7.3 keyring/23.4.1 rfc3986/1.4.0 colorama/0.4.4 CPython/3.6.13

File hashes

Hashes for alvenirclient-0.0.2.tar.gz
Algorithm Hash digest
SHA256 fb90c1766634855737db5752e22ff04dbff3d30d2004db13d4429b3dcb91f999
MD5 7ca28c2537fdf946a1696538ffbb8513
BLAKE2b-256 469281ef2ad407eb5319ef6baf2813cff80045d0684e655a4cef07b6a34bc59f

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page