This project contains individual modules created for the ToHyve-project. All modules are also provided as docker containers for higher usability.
The individual models encompass:
- Multilingual Automatic Speech Recognition (ASR) (automatic_speech_recognition)
- Multilingual Bitext Alignment (DFKI_NLU_DOCKER/bi_text_alignment)
- Multilingual Text to Text translation / Text to Speech translation (DFKI_NLU_DOCKER/machine_translator)
- Multilingual Text To Speech conversion (TTS) (text_to_speech_conversion)
- Multilingual ASR (Video (without caption) -> Extract Audio -> Extract Text from Audio -> Translate that Text into English, German -> Add Translated Texts as a caption of inserted Video) (DFKI_NLU_DOCKER/dfki_asr)
- Bite Cutter: Tool to support the TTS service. It accepts longer input than the TTS service does, cuts them in smaller pieces and resends them to the service. Functions as a wrapper container.
To use the stream capabilities of ASR and then integrate with other services we created a small python script request_stream_demo.py.
To use it:
python -m request_stream_demo
N.B: Plase open the script and change the variable values with your actual values and to use TTS audio don't forget to uncomment tts_audio_encodings(line 85). The TTS encodings are commented out for better visibility.