Spleeter
Free.ai (self-hosted)
·
audio_separate
·
~1000 Token track
Spleeter bụ a steém n'eziga móòdù e mepụtara site na Deezer. Ọrụ na DJ/production workflows, karaoke extraction.. Self-hosted na Free.ai GPUs - na-agba ọsọ n'efu megide ụbọchị gị token pool (1,000 tokens N'otu ụzọ). E wepụtara ya n'okpuru MIT - iji ahịa ekwenyela na Free.ai.
Ajụjụ ndị a jụrụkarị
Spleeter na-ewepụ ụda-n'ime-n'ime n'ime stiimi - vocals, drums, bass, na ndị ọzọ n'ime n'ime - ọbụla dịka faịlụ ụda ya. Bubata MP3 mọọbụ WAV na Spleeter na-ewepụ 4 n'ime-n'ime stiimi faịlụ ị nwere ike ịgbanwe ma ọ bụ kwụsị.
Nchedo na-emechi emechi na stiọdi tráìkì ndị a na-emechi: vokal na instrumentals dị iche iche na obere bleéị̀, na drums na bass na-apụta n'ụzọ dị ọcha. Oge ochie, ndụ ma ọ bụ n'ụzọ dị ukwuu a na-akpụchaghị n'ime ntọhapụ dị iche iche na n'etiti steelụ ndị ọzọ.
Stems anọ: vocals, drums, bass na ndị ọzọ. Intọlitelụ n'ime n'ime (nke niile n'agbanyeghị vocals) dịkwa ka ọbụna otu-pịa-ọnụọgụgụ.
MP3, WAV, FLAC, M4A, OGG ruo 50 MB. Stereo init na-enye ihenhọrọ ndị dị ọcha; mono a na-ebuli ya n'ime stereo tupu a na-ejikwa ya.
Spleeter na-agba ọsọ na GPUs anyị megide ụbọchị gị n'efu pool mbụ; $ 5 → 200,000 akwụ ụgwọ tokens. About ~ 1,000 tokens kwa nkeji nke ụda.
Ee - wepụta stims, wee bubata "drums + bass + other" nke e ji jikọtara ọnụ maka instọlụ, mọọbụ bubata "vocals" naanị maka a-cappella. Fomu n'elu nwere otu-pịa "instọọlụ" ọbjektị.
Ee - ngbanwe n'etiti na-agbanweghị oge mọọbụ n'elu. Steelụ na-ejikọta fréèmù-n'ime fréèmù na isi mmalite ka ha wee banye n'otu n'otu.
Ee - /batch/ na-anabata nsomebefaịlụ nke faịlụ ụda. Ogo ọbụla (4 stiimi n'otu isi) na-abanye na /account/?tab=history. Maka nchekwa nsomebefaịlụ-tree jiri API.
Ee. POST ụda gị na /v1/music/separate/ na ị ga-enweta otu mbubata URL kwa stiim azụ. /api/ nwere nlekọta zuru ezu, gụnyere model id iji zitere.
Ichọrọ ikikembanye na ụda isi (n'ime gị, ihenhọrọ nke ikikembanye, ma ọ bụ n'ime egwu na ikikembanye remix). Free.ai agaghị enye ikikembanye na ihenhọrọ ị na-enweghị; ịgwakọta ndepụta egwu pop na-egbochi ịdepụta stiimi.
WAV site na difọ́ọ̀ltụ̀ (enweghị nkwụsị, nọ n'ọrụ iji megharịa). MP3 dị na nhọrọ ụkpụrụedemede maka mbido n'okpuru.
Real-time factor bụ ihe dịka 0.1-0.3× - 4-minute song na-ewepụ n'ime 30-80 sekọnd. Jiri bọtịn n'oche iji mechie táàbụ̀ ahụ.