ä¿¡ããããªãã»ã©ãªã¢ã«ãªé³å£°ãçæ
Fish Audio S2ã詊ãS2ã®ç¹åŸŽ
衚çŸåãã¹ããŒãããªãŒãã³ãã¹ããã¹ãŠããŒããã远æ±ããé³å£°AIã
è¶ äœé å»¶
150msæªæºã®é«éã¬ã¹ãã³ã¹ã§ããªã¢ã«ã¿ã€ã ãªå¯Ÿè©±åAIãã©ã€ãå¹ãæ¿ããã€ã³ã¿ã©ã¯ãã£ããªé³å£°ã¢ããªãå®çŸãé³è³ªãèœãšãããšãªããåçšç°å¢ã§ãå®å¿ããŠäœ¿ããå§åçãªããã©ãŒãã³ã¹ãæäŸããŸãã

<150ms
150msæªæºã®å¿çé床ã«ããããªã¢ã«ã¿ã€ã äŒè©±AIãã©ã€ã广¿ãã€ã³ã¿ã©ã¯ãã£ããªé³å£°ã¢ããªãå®çŸãå質ã劥åããããšãªããæ¬çªéçšã«èãããå§å·»ã®ããã©ãŒãã³ã¹ãçºæ®ããŸãã
ããã¹ãæç€ºã ãã§ãææ ã声ã®è¡šæ ãèªç±èªåšã«ã³ã³ãããŒã«ãç¬ã声ããããããããæ¯ãªã©ããããã衚çŸãç°¡åã«è¿œå ã§ããŸããããã«ãäžåºŠã®é³å£°çæã§è€æ°ã®è©±è ãã¹ã ãŒãºã«åãæ¿ããèªç¶ãªäŒè©±ã·ãŒã³ãäœæå¯èœã§ãã

<|speaker:1|> [giggles]
å®å šãªãŒãã³ãœãŒã¹
æšè«ã³ãŒããšã¢ãã«éã¿ã®äž¡æ¹ãå®å šãªãŒãã³ãœãŒã¹ã§å ¬éãèªç€Ÿã€ã³ãã©ã§ã®éçšãç¬èªããŒã¿ã«ãããã¡ã€ã³ãã¥ãŒãã³ã°ããã³ããŒããã¯ã€ã³ã®ãªãèªç±ãªã·ã¹ãã çµ±åãå¯èœã§ããé«ãéææ§ãšãã³ãã¥ããã£äž»å°ã®ã€ãããŒã·ã§ã³ãç®æããŠéçºãããŠããŸãã

Built with SGLang
Fish Audio S2 APIã§æ§ç¯
80以äžã®èšèªã«å¯Ÿå¿ããææ ãæŒåºæç€ºãè€æ°è©±è ã³ã³ãããŒã«ãåãããªã¢ã«ãªé³å£°ãçæããŸãã

from fishaudio import FishAudio
from fishaudio.utils import save
# Initialize with your API key
client = FishAudio(api_key="your_api_key_here")
# Generate speech
audio = client.tts.convert(text="Fish Audio S2 is the best voice AI model.", model="s2-pro")
save(audio, "welcome.mp3")ãããã質å
Fish Audio S2 Proã¯ãé»åŸãšææ ã®ãã现ãããªã€ã³ã©ã€ã³å¶åŸ¡ãåããæå 端ã®ããã¹ãèªã¿äžãã¢ãã«ã§ãã80以äžã®èšèªã«ããã1,000äžæé以äžã®é³å£°ããŒã¿ã§åŠç¿ããã匷ååŠç¿ã¢ã©ã€ã³ã¡ã³ããšãã¥ã¢ã«èªå·±ååž°ïŒDual-ARïŒã¢ãŒããã¯ãã£ãçµã¿åãããŠããŸãã40åãã©ã¡ãŒã¿ã®Slow ARãæå³äºæž¬ãã4åãã©ã¡ãŒã¿ã®Fast ARãé³é¿ã®è©³çްåŠçãæ åœãã¢ãã«éã¿ããã¡ã€ã³ãã¥ãŒãã³ã°ã³ãŒããããã³SGLangããŒã¹ã®ã¹ããªãŒãã³ã°æšè«ãšã³ãžã³ãå ¬éããŠããŸãã
S2 Proã¯ã[tag] æ§æã䜿çšããŠããã¹ãå ã«èªç¶èšèªã®æç€ºãçŽæ¥åã蟌ãããšã§ãé³å£°çæã®ãã³ãã€ã³ããªå¶åŸ¡ãå®çŸããŸãããããããå®çŸ©ãããã¿ã°ã»ããã«äŸåããã[whisper in small voice]ã[professional broadcast tone]ã[pitch up] ãªã©ã®èªç±åœ¢åŒãªããã¹ã衚çŸãåãå ¥ããåèªã¬ãã«ã§ã®èªç±åºŠã®é«ã衚çŸå¶åŸ¡ãå¯èœã«ããŸãã[pause]ã[emphasis]ã[laughing]ã[excited]ã[whisper]ã[singing] ãªã©ã15,000çš®é¡ä»¥äžã®ã¿ã°ããµããŒããããŠããŸãã
1æã®NVIDIA H200 GPUäžã§ãS2 Proã¯ãªã¢ã«ã¿ã€ã ãã¡ã¯ã¿ãŒïŒRTFïŒ0.195ãæåã®é³å£°ãåºåããããŸã§ã®æéçŽ100msãæ¯ç§3,000以äžã®é³é¿ããŒã¯ã³ã¹ã«ãŒããããéæãã€ã€ãRTFã0.5以äžã«ç¶æããŸããSGLangããŒã¹ã®æšè«ãšã³ãžã³ã¯ãé£ç¶ãããåŠçãPaged KV CacheãCUDAã°ã©ããªãã¬ã€ãRadixAttentionããŒã¹ã®ãã¬ãã£ãã¯ã¹ãã£ãã·ã³ã°ãªã©ãããããLLMãã€ãã£ããªãµãŒãã³ã°æé©åãç¶æ¿ããŠããŸãã
S2 Proã¯80以äžã®èšèªããµããŒãããŠããŸããTier 1èšèªïŒæé«å質ïŒã«ã¯æ¥æ¬èªãè±èªãäžåœèªãå«ãŸããŸããTier 2èšèªã«ã¯éåœèªãã¹ãã€ã³èªããã«ãã¬ã«èªãã¢ã©ãã¢èªããã·ã¢èªããã©ã³ã¹èªããã€ãèªãå«ãŸããŸããã¹ãŠã§ãŒãã³èªãã€ã¿ãªã¢èªããã«ã³èªããªã©ã³ãèªããã³ãã£ãŒèªãã¿ã€èªããããã èªãªã©ã察å¿ããŠããŸãã
S2 Proã¯Fish Audio Research Licenseã®äžã§ã©ã€ã»ã³ã¹ãããŠããŸããç ç©¶ããã³éåçšå©çšã¯ç¡æã§èš±å¯ãããŠããŸããåçšå©çšã«ã¯Fish Audioããã®å¥éã©ã€ã»ã³ã¹ãå¿ èŠã§ãã詳现ã¯business@fish.audioãŸã§ãåãåãããã ããã
