Guide-yakabatana mibvunzo · Pakati Nhanho

Mubvunzo Wekushandura-Kutaura-Kutaura

Inotarisa kududzira kwakadhindwa uye kwakananga kwekutaura, discrete acoustic units, kuyerera latency, kuchengetedza inzwi uye kushushikana kwemvumo.

Nzira dzekutungamira dzinoenderanaDudziro Yekutaura Kutaura
Mubvunzo 1 ye 8

Shanduro yeKutaura-ku-Kutaura (S2ST) inoita sei?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Shanduro yeKutaura-ku-Kutaura (S2ST) inoita sei?
    • Inoshandura kutaura mumutauro mumwe kuva mazwi anotaurika mune mumwe mutauro
    • Inongonyora odhiyo kuita zvinyorwa
    • Kushandura magwaro akanyorwa kuenda kumaPDF
    • Inogadzira mimhanzi yekumashure
  2. Ndeapi matanho matatu eiyo classic cascaded S2ST system?
    • Rekodha, chengeta, bvisa
    • Kuziva kutaura, kushandura nemuchina, kubva-ku-kutaura-mavara
    • Encrypt, compress, tumira
    • Teerera, furatira, dzokorora
  3. Chii chakakosha mukana weakananga (kumagumo-kusvika-kumagumo) S2ST pamusoro pecascade masisitimu?
    • Inogara ichida mamwe matanho emavara
    • Haikwanise kushandura zvachose
    • Inogona kuderedza latency uye kuchengetedza zviri nani inzwi uye prosody
    • Inongoshanda pamitauro miviri
  4. Ndeipi Meta system inozivikanwa nekushandura mitauro ingangoita 100?
    • ChatGPT
    • Midjourney
    • Stable Diffusion
    • SeamlessM4T
  5. Sei shanduro yenguva chaiyo yekutepfenyura ichinyanya kunetsa?
    • Sistimu inofanirwa kusarudza nguva yekushandura mutsara usati wapera
    • Audio haigone kufambiswa
    • Mitauro yose inorongana mazwi akafanana
    • Hazvidi mhando
  6. Kuitira hanya kupi kuri kubuda kunobva mukuchengetedza inzwi romukurukuri mukushandura?
    • Mafaira anova madiki
    • Mvumo uye deepfake/voice-cloning kushandisa zvisizvo
    • Shanduro inova isingagoneki
    • Audio inorasa ruzha rwese
  7. What representation may a direct speech-to-speech system use to connect source audio to target audio?
    • A Slurm job identifier
    • A table of image pixels
    • Discrete acoustic units that serve as intermediate speech tokens
    • A hand-written class label for every word
  8. Why can errors accumulate in a cascaded ASR → translation → TTS pipeline?
    • Each stage uses the same model weights
    • TTS always corrects the transcript
    • The stages run in parallel on the same audio
    • An upstream transcription or translation error becomes input to later stages