Use OpenAI Whisper to recognize human voices from audio files.
Read the information of the audio file from the audio_file field and call whisper for recognition. Finally, the language of the recognized voice is returned from the language field; the full text is returned from the text field; the segmented recognized text and other related information are returned from the segments field.
The upstream of the model field must be a Whisper model shared block, and the received model will be used to recognize human voices in audio files.
The word_timestamps field indicates whether the segments segment accurately marks the time range of each word.
The prompt field will be submitted to the model as a prompt.
Configure and load the Whisper model, which is returned in the model field and used by the downstream Whisper block.
The model field selects the type of model to be loaded.
device indicates whether CUDA is used, otherwise the CPU is used.
dir_path The folder from which to read the model (if the model does not exist, it will be downloaded to this folder). If not filled in, it will be the default folder of Whisper.
Read the text segment information of the upstream Whisper block from the segments field, convert it to SRT format text, and return it from the srt field.