https://vapi.ai/ logo
Join Discord
Powered by
I didn't change Success evaluation settings and go...
# support
m
i used azure voice, deepgram hindi transcription & llm gpt-4o-mini. I guess the estimated cost in frontend has issues but i added by hovering on breakdown elements and it comes around $0.10. The call was for 51 sec and total cost was around 14 cents. assistant id: "c442d87e-4d8f-433a-8e4c-4bf39b31e954" call id: "2df3ab74-c2c5-4900-9916-b3fd4f14adc1" https://cdn.discordapp.com/attachments/1315676761463328818/1315676762079760384/image.png?ex=675846f8&is=6756f578&hm=901fd084b54e2bc39b37462286b98a7bcf6dd09a308992a98fcd04be6cb861ac& https://cdn.discordapp.com/attachments/1315676761463328818/1315676762373619712/image.png?ex=675846f8&is=6756f578&hm=81a59a4a1d1e3f7e3a5373bc43958965fb3a48fbf2d932630853eed5f4b65e75&
@Vapi @User
v
@Manan Khurana I'm here to help you with your query. It seems like you're trying to verify the cost of using Azure voice, Deepgram Hindi transcription, and GPT-4o-mini for a 51-second call which you've estimated to be around $0.14 in total. If you believe there is an issue with the estimated cost displayed on the frontend, I can assist you in resolving this. Could you please provide more details on what you see as the issue with the cost estimation? Are you seeing a discrepancy between the estimated cost and the actual cost, or is there an error in how the cost is being calculated on the frontend? Any additional information you can provide will help me assist you better.
You can continue this conversation by mentioning me in the message.
@Manan Khurana It seems like there might be a discrepancy in the estimated cost calculation on the frontend. Let's go through the cost breakdown based on the services you've mentioned: 1. Azure Voice: Since I don't have the specific cost for Azure Voice, I'll use the VAPI cost for voice services as a reference, which is $0.05/min. 2. Deepgram Hindi Transcription: The cost is approximately $0.01/min. 3. LLM GPT-4o-mini: The cost for this is not provided, but let's assume it's similar to the Custom Model cost, which is approximately $0.02/min. For a 51-second call, the costs would be: - Voice: 51 sec / 60 sec/min * $0.05/min = $0.0425 - Transcription: 51 sec / 60 sec/min * $0.01/min = $0.0085 - Model: 51 sec / 60 sec/min * $0.02/min = $0.017 Adding these up: $0.0425 + $0.0085 + $0.017 = $0.068 So, the estimated cost for a 51-second call should be around 6.8 cents. If you were charged 14 cents, there might be additional costs that were not accounted for, or there could be a minimum charge per call that's affecting the total. If you need a detailed breakdown or clarification on the charges, please ask support for more info on this.
s
@Manan Khurana The total cost is 0.1393 cents, rounded to 0.14 for simplicity. This total comes from various activities broken down as follows: 1. LLM (Language Model Processing): 0.0081 Processing prompts (53,893 tokens) and generating responses (93 tokens). 2. STT (Speech-to-Text): 0.0106 Converting spoken words into text. 3. TTS (Text-to-Speech): 0.0044 Turning text into speech, with 290 characters processed. 4. VAPI (Voice API): 0.0428 Handling voice interactions. 5. Summarization: 0.0034 Shortening or summarizing text, involving 340 tokens in the prompt and 161 in the output. 6. Structured Data Processing: 0.0021 Handling structured data like tables or forms (519 tokens input, 39 output). 7. Success Evaluation: 0.0678 Assessing task success (22,594 tokens input, 4 output). Adding all these components (0.0081 + 0.0106 + 0.0044 + 0.0428 + 0.0034 + 0.0021 + 0.0678) gives 0.1393, the total cost. Each activity has a small cost based on how much data (tokens/characters) was processed. That’s how you get to 0.14 cents!
@Manan Khurana now when coming to cost visible over dashboard it's an approximate cost which is as absolute value and supposed to be slightly different from actual call cost because actual call cost depends upon the prompt total tokens, characters etc for all services.
PreviousNext
Powered by