i think youre judging it on the wrong axis tbh. its not pitched as realtime, its pitched as quality. for a podcast or audiobook the latency genuinely doesnt matter and the output is some of the best ive heard

was about to agree the latency makes it useless then i realized i was streaming chunks wrong. fixed my buffering and its way better. partly a me problem

built for batch not live is just a fancy way of saying its slow and im at peace with it lol

whats the cost per minute of audio. cant find clear pricing anywhere

genuine question, when you say latency is rough whats the actual number youre seeing? im trying to figure out if its the model inference or the network hop. i got around 1.4s end to end from europe and i suspected most of it was distance not compute. if its consistently slow regardless of region thats a different story and worth flagging to them

i have to push back on the framing a little. realtime latency is a real requirement for a huge slice of voice use cases, agents, accessibility tools, live translation. saying just use it for batch quietly writes off the hardest and most valuable part of the market. if vocence wants to be more than a nice audiobook tool they do need to solve the live path, and pretending its fine as is does them no favors 🌙

how many languages does it actually do well though? saw the list but lists always lie. anyone tested the non english ones

does it do voice cloning or just preset voices. thats the dealbreaker for me not the latency

the VOICE QUALITY on this is unreal tho 🔥 like i played it for my mom and she thought it was a real person. who cares about a half second when it sounds THAT good

嫌慢的可以去用机器人音啊 人家这音质值得等 一秒钟你都等不了 哈哈 🎤

started this thread thinking the latency was a dealbreaker but after using it for a week ive come around 😅 for what i actually do it never mattered once. quality wins 🎧

行吧我收回 自己拿来做实时客服确实卡 batch场景是真香 实时还得练 别杠了我自己打脸

i made it read my grocery list in the most dramatic voice and honestly worth the latency for that alone lol

still think the realtime complaint misses the point but i hear you that they should be clearer about it on the page. people keep showing up expecting live and bouncing

ok i originally thought noir was nitpicking but i tried wiring it into a live agent and yeah the lag kills it. for batch its lovely, for live its a no. changed my mind

under a second tts in 2026 and people act like its owed to them. the audacity. it sounds like a human, give it the half second ⚡