I was checking within 11Labs, and I already purchased the Creator plan, cloned my voice, and everything is set up. I also connected the API to BAPI and assigned that voice to my assistant, and it’s working fine. The issue is that I’ve noticed within 11Labs that each time I use that voice, my credits decrease. I consume about 1,000 credits per call, and from what I understand, the Creator plan in 11Labs provides 100,000 credits, which will allow me to make approximately 100 calls. So, my question is: when I use voices provided by BAPI from 11Labs, am I consuming additional credits? In other words, does the cost not get added separately to 11Labs? However, when I use my cloned voice, I understand I also need to pay the additional cost within 11Labs. Is that correct? Could someone explain how this works and clarify the cost differences between using the voices provided by BAPI within 11Labs versus using my own cloned voice in 11Labs?

@Paolo I'm glad to hear that you've successfully set up your cloned voice with BAPI and that it's working well for you. Regarding your query about credit consumption and costs, here's how it works: When you use the voices provided by BAPI, including your own cloned voice from Elevenlabs (11Labs), you are indeed consuming credits from your Elevenlabs account. Each API call to generate speech with your cloned voice deducts credits from your balance based on the length of the generated audio and the specific pricing plan you are on. With the Creator plan, you mentioned that you receive 100,000 credits. If you're consuming about 1,000 credits per call, it means you can make approximately 100 calls, as you've calculated. This consumption rate is part of the service provided by Elevenlabs, and it's separate from any other costs or credits you might have with BAPI. The cost difference between using standard BAPI voices and your own cloned voice from Elevenlabs typically lies in the credit consumption rate. Standard voices might have a different rate of credit consumption compared to a cloned voice, which could be higher due to the personalized nature of the service. To summarize, using any voice through the Elevenlabs API, whether a standard voice or your cloned voice, will consume credits from your Elevenlabs account. The rate at which credits are consumed depends on the pricing plan you're on and the type of voice you're using. If you need a detailed breakdown of credit consumption for different types of voices or further clarification on pricing, please ask support for more info on this. They can provide you with the most accurate and up-to-date information tailored to your specific use case and plan.

Pero mi pregunta entonces es, ¿por qué cuando ni siquiera incluyo la API de Levenlabs, igual puedo utilizar voces que son parte de Levenlabs? O no sé si es que, ¿cómo funciona eso? Porque yo sin tener una cuenta de Levenlabs ya podía utilizar un par de voces. Entonces, ¿por qué cuando utilizo una voz, digamos, importada desde mi cuenta de Levenlabs, me cobran adicionalmente esos créditos, y si es que utilizo las voces que proporcionaba API, no hay ese consumo de créditos?

If you can find a voice in the Voice Library instead of using your own voice that would be significantly cheaper for you. Unfortunately, the only way to use your own custom 11labs voice is using your API key to access it - this will then switch you to using your own credits. Stick with VAPI's super low rates if you can.

Thanks a lot chris!

@Paolo First, it is not BAPI, it's VAPI. Regarding the costs, as @Chris Laupama has answered, I believe your initial doubt is resolved - if not, please let me know. As for using other telephony providers, you can use them within VAPI through SIP Trunking. You'll need to check with your telephony provider if they support this feature. If they do, then you can use any number with VAPI. Let me know if your telephony providers supports SIP Trunking then I will share the next process.

all right i will let you know as soon i have. response from my telephony providers

1. Eleven Labs is currently the best in the market. If possible, use the Turbo V2 and multilingual models instead of the Flash model for now, as it's still new and may have some issues. In terms of latency, all models perform well except when compared to Flash. 2. For the Deepgram transcriber, set its endpoint timing to 300ms and waitSeconds to 0.8 to better capture words. The Stop Speaking Plan works well. Regarding your questions about "number providers to minimize response times" and "parameters for optimal response time and audio quality" - could you please provide more context about what specific aspects you'd like to know about?