AI Voice Sales Assistant
Budget: €30 – €250 EUR
I am building a desktop-based voice assistant that can jump on live, one-to-one calls with prospects in Romanian, handle their sales-related questions, and keep the conversation flowing naturally through the ChatGPT API. The core flow is simple: the app dials (or receives) a call, converts speech to text, lets ChatGPT generate the response, and returns it to the caller via realistic text-to-speech—all fast enough to feel like human conversation.
Key expectations
• Desktop application (Windows is the immediate need; macOS portability is a plus)
• Fluent, idiomatic Romanian output and recognition
• Conversation logic driven by the ChatGPT API with the ability to inject my own product information and sales prompts
• Stable audio pipeline: real-time STT, low-latency TTS, and smooth voice hand-off between user and assistant
• Call control UI showing transcript in real time so an operator can watch or intervene if needed
Deliverables
1. Executable desktop app and documented source code
2. Setup guide covering required API keys (OpenAI, telephony provider of your choice)
3. Short video demo proving a live call in Romanian
I’m ready to test quickly and iterate, so clean, modular code is important. If you’ve already wired voice, telephony, and GPT together, I’d love to see it.
Key expectations
• Desktop application (Windows is the immediate need; macOS portability is a plus)
• Fluent, idiomatic Romanian output and recognition
• Conversation logic driven by the ChatGPT API with the ability to inject my own product information and sales prompts
• Stable audio pipeline: real-time STT, low-latency TTS, and smooth voice hand-off between user and assistant
• Call control UI showing transcript in real time so an operator can watch or intervene if needed
Deliverables
1. Executable desktop app and documented source code
2. Setup guide covering required API keys (OpenAI, telephony provider of your choice)
3. Short video demo proving a live call in Romanian
I’m ready to test quickly and iterate, so clean, modular code is important. If you’ve already wired voice, telephony, and GPT together, I’d love to see it.