1. DBMake loads database scheme(.cds) and generates spliting configuration(.sconf).
2. The abridged spliting configuration(.sconf) is loaded into WSplit along with raw data(.wav). Spliting configuration with time intervals(.sconf) is generated.
3. The time-shifted spliting configuration(.sconf) is sent to WSplit. Splited sample(.wsp) and sample configuration(.wconf) is generated.
4. The frequency-modified sample configuration(.wconf) is sent to DBMake. Cybervoice Dictionary(.cdt) is generated.
5. The sample configuration and splited sample are sent to Wpp. Splited sample(.wsp) is regenerated after preprocessing.
6. Based on preprocessed splited sample(.wsp), the average formant information stored in database scheme(.cds), and sample configuration(.wconf), CMin generates database fragments(.cvdb).
7. Database fragments(.cvdb) are sent to DBPack, which packs them into the final singer database(.cdb).
[Debuging]
a. Database fragments(.cvdb) are sent to C2Wav, waves with periodic marks(.wav) are generated. Further checking works can be done.