20190330 Song junyeong(Leader)
20190422 Yoon sunmoon
20190516 Lee chanryeol
The model is divided into six stages.
- Extracting song information: Using ACRCcloud and GPT API
- Original lyrics extraction: Website crawling
- Separate input music file: Use Spletter
- Lyrics Translation: Based on GPT-4o model
- Create music: Use suno.ai or ACEStudio
- If suno.ai, use api so end-to-end
- If ACEStudio, use translated lyrics and separated music, allocate syllables into note manually(example results are 가시_ACE)
- Voice conversion: Use AICoverGen
- Install libraries using requirements.txt
- In lyrics_extract.py, and lyrics_translation.py, write your gpt api key in "your_api_key"
- Input mp3 (or m4a) file in dataflow/music_original
- Input voice model in AICoverGen/rvc_model
- sh Adopting.sh {file_name} {artist_name} {language pair} {model}
- file_name: name of mp3(mp4) file in dataflow/music_original
- artist_name: name of voice model to conversion
- language pair: original language, translating language pair, only three characters (ex. kor,eng)
- model: GPT-api model (ex. gpt-4o, gpt-4, gpt-3.5-turbo)
ex) sh Adopting.sh weekend.m4a Taeyeon kor,eng gpt-4o