- Ai will arbitrarily add periods to your sentences.
- The program will force line breaks according to the period.
Then your content swings away.
It wasn’t like this before
upload MP3 & SRT
- The sentences inside the course are different from my SRT.
upload SRT only
- The sentences inside the course are different from my SRT too.
My thoughts:
Don’t use use AI to preprocess my srt
Don’t use Whisper to process my audio
I’ll handle it myself
here’s a good one:
{
"faster_whisper_config": {
"model_path": " ",
"local_files_only": true,
"download_root": " ",
"device": "cuda:0",
"compute_type": "float16",
"transcribe_options": {
"language": "de",
"task": "transcribe",
"beam_size": 10,
"best_of": 5,
"patience": 1.0,
"length_penalty": 2.0,
"temperature": [0.2, 0.4, 0.6, 0.8, 1.0],
"compression_ratio_threshold": 2.4,
"log_prob_threshold": -1.0,
"no_speech_threshold": 0.0,
"word_timestamps": true,
"suppress_blank": true,
"suppress_tokens": [-1],
"repetition_penalty": 1.0,
"no_repeat_ngram_size": 0,
"prompt_reset_on_temperature": 0.5,
"chunk_length": 30,
"initial_prompt": "",
"prefix": "",
"hotwords": "This is a audio, please detect all inflections and use more punctuation to enrich the emotion, such as ellipses, question marks, full stops, exclamation points, dashes(...) ? !)"
},
"thread_num": 24,
"num_worker": 1
}
}