Adding onto this, does the length of the
# general-englisha
Alex
05/07/2025, 2:18 PMAdding onto this, does the length of the script also make the assistant slower? I want mine to ask my customers some specific questions so it's kinda long you could say
p
Pigeon
05/07/2025, 2:23 PMCheck your call logs for [INFO] tags, these show what take how much time
https://cdn.discordapp.com/attachments/1369679792231940197/1369680955584942090/image.png?ex=681cbe4b&is=681b6ccb&hm=5083ac566a7fbcb75f975d3bedf3fbd9509e8eb5f47ce2d927c5d76f6f0b5705&
Pigeon
05/07/2025, 2:24 PMYes, longer prompts can slow you down a bit
Pigeon
05/07/2025, 2:25 PMShow me some INFO logs and Ill try to help you
a
Alex
05/07/2025, 2:27 PMAlex
05/07/2025, 2:28 PMhttps://i.imgur.com/SrBASSj.png▾
p
Pigeon
05/07/2025, 2:28 PMYes, dont use livekit with non english 🙂
p
Pigeon
05/07/2025, 2:28 PMYour voice latency is quite large
Pigeon
05/07/2025, 2:28 PMWhat model are you using?
Pigeon
05/07/2025, 2:28 PMSTT/Voice model
a
Alex
05/07/2025, 2:29 PMi'm using open ai with gpt 4.1
p
Pigeon
05/07/2025, 2:30 PMPigeon
05/07/2025, 2:30 PMOr is it openai here too?
p
Pigeon
05/07/2025, 2:31 PMHmm, maybe try 11labs flash 2.5 model, youll have to find german voice in https://elevenlabs.io/app/voice-lab
Pigeon
05/07/2025, 2:31 PMWait, Ill give you ID to try
a
Alex
05/07/2025, 2:32 PMYeah i need a really pleasant sounding german female voice, that was one of the few good options i found
p
Pigeon
05/07/2025, 2:33 PMdCnu06FiOZma2KVNUoPZZQFCSsF1tIcjtMZJ6VCAa
Alex
05/07/2025, 2:33 PMAight, sec 🙏
p
Pigeon
05/07/2025, 2:33 PMIf speed is good you can brows many voices and chose what you like
Pigeon
05/07/2025, 2:33 PMa
Alex
05/07/2025, 2:37 PMI've tried both voices, but the answering speed seems to be the same
https://cdn.discordapp.com/attachments/1369679792231940197/1369684537822413050/rhAG2vW.png?ex=681cc1a1&is=681b7021&hm=99c513ddcd94d6c6d6ff01ccb5709378d07078b28cefd221ea5c089f489cefce&
Alex
05/07/2025, 2:37 PMWhenever i say something, the assistant needs a good 2-5 seconds to answer
p
Pigeon
05/07/2025, 2:40 PMWell, if you're in Europe, add ~400ms of network latency
How do you get 0ms on transcriber?
Pigeon
05/07/2025, 2:40 PMOh wait, mine says 0ms too
a
Alex
05/07/2025, 2:41 PM1-2 seconds would be completely fine, but mine sometimes needs like up to 5 seconds
p
Pigeon
05/07/2025, 2:41 PMYeah, so try to find [INFO] for 5s pause
Pigeon
05/07/2025, 2:42 PMMaybe its LLM, OpenAI llms are not at all stable in response times
a
Alex
05/07/2025, 2:43 PMAny suggestions i could check out?
p
Pigeon
05/07/2025, 2:46 PMI use groq, they are faster, more stable, but a lot dumber compared to OpenAI models
Pigeon
05/07/2025, 2:46 PMMaybe try 4.1 mini, 4.1 nano
p
Pigeon
05/07/2025, 3:00 PMAre you testing through the phone or web call?
a
Alex
05/07/2025, 3:02 PMI tried all methods basically. Web call, demo link and call request via make.com automation
p
Pigeon
05/07/2025, 3:03 PMSo what are the times in
[INFO]a
Alex
05/07/2025, 3:10 PMThe timing in the log looks fine, but the call itself sounds much more delayed
Alex
05/07/2025, 3:10 PMAlex
05/07/2025, 3:11 PMhttps://cdn.discordapp.com/attachments/1369679792231940197/1369693125878222972/at5bbWF.png?ex=681cc9a0&is=681b7820&hm=a6c646262f68ba98122a04caa0edd4ab48a1e661094e3a2d4b812f6b83cda126&
https://cdn.discordapp.com/attachments/1369679792231940197/1369693126075088896/kR1kJvk.png?ex=681cc9a0&is=681b7820&hm=749ef6b74c457b4f07361f91b6ca4d2399f604c7c04845bd36ad463c9597d8a2&
Alex
05/07/2025, 3:11 PMSadly i have no clue what those mean yet
2 Views