Demucs
Free.ai (self-hosted)
·
audio_separate
·
~1000 Token kwa track
Demucs bụ a steém separator móòdù e mepụtara site na Meta. Ọrụ na Vocal/stem separation. 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 azụmahịa ekwenyela na Free.ai.
Jiri site na API
OpenAI-na-akpaghị aka REST API. Kewapụta kii nakwa kpọọ móòdù a n'ime sekọnd.
curl -X POST https://api.free.ai/v1/audio/separate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"demucs","prompt":"your prompt here"}'Ajụjụ ndị a jụrụkarị
Demucs na-ewepụ ụda ọfụụ n'ime stims - vocals, drums, bass, nakwa ndị ọzọ n'ime n'ime - otu ọbụla dị ka faịlụ ụda ya. Bipụta MP3 mọọbụ WAV na Demucs na-ewepụ 4 faịlụ ọfụụ ị nwere ike ịgbanwe ma ọ bụ wepụ.
Demucs-family separator bụ nke dị n'ọdịnihu - vocals bụ nke dị nso na studio-quality na mbipụta oge ọfụụ, drums na bass bụ nke dị ọcha. Otú ọ dị, ma ọ bụ ihenhọrọ nke a na-akpụgharị n'ụzọ dị ukwuu na-enye nkwụsị nke na-akpụgharị.
Standard 4-stem: vocals, drums, bass, ndị ọzọ. Otú ụfọdụ model si enyere 6-stem (ịgbakwunye piano + guitar n'otu) — họrọ ụdị n'ime usoro ahụ.
MP3, WAV, FLAC, M4A, OGG ruo 50 MB. Stereo input na-enye nsonaazụ kacha mma; mono na-akpụzi ya ka ọ bụrụ stereo tupu a na-arụ ya.
Demucs na-agba ọsọ na GPUs anyị n'ihu n'ụbọchị gị n'efu; $5 → 200,000 tokens paid. About ~1,000 tokens per minute of audio.
Yabụ - wepụta stims, wee bubata "drums + bass + ndị ọzọ" stitched n'otu maka instrumental, mọọbụ bubata "vocals" naanị maka a-cappella. Fọmu n'elu nwere otu-pịa "instrumental" ọbjektị.
Yabụ - ngbanwe n'etiti na-agbanweghị oge mọọbụ n'elu. Stems na-ejikọta fréèmù-n'ime frèèmù na isi mmalite ka ha wee gbanwee n'ụzọ dị ọcha.
Ee - /batch/ na-anabata nsomebefaịlụ nke faịlụ ụda. Ogo ọbụla (4 stims kwa isi) na-abanye na /account/?tab=history. Maka nchekwa nsomebefaịlụ-tree jiri API.
Yabụ - POST ụda gị na /v1/music/separate/ na model="Demucs". Na-eziga 4 mbipụta URLs (otu n'otu na stiim). /api/ nwere ndesịta zuru ezu.
Ichọrọ ikikembanye na ụda isi (n'aka gị, ihenhọrọ nke nwere ikikembanye, mọọbụ abụ na ikikembanye remix). Free.ai agaghị enye ikikembanye na ihenhọrọ ịnweghị ikikembanye; ịgwakọta abụ pop cháàs na-egbochi ịgbapụta steelụ.
WAV site na difọ́ọ̀ltụ̀ (enweghị nkwụsị, nọ n'ọrụ iji megharịa). MP3 dị na onyehọrọ ụhara maka mbipụta n'okpuru.
Oge-ezighi ezi faktor bụ 0.1-0.3× - 4-minuite n'akụkọ ihe mere eme na-ewepụ n'ime sekọnd 30-80. Jiri bọtịn n'okporo ụzọ mechie táàbụ̀ ahụ.