Puheen synteesi Jobs

My pathway already performs smoothly in text, but the moment we switch to voice a few critical problems appear. The system starts throwing spelling errors, misunderstands certain words, and the overall exchange stops sounding human-like. I need these pain points fixed so the voice experience feels just as polished as the text one. What I want improved • Spelling accuracy and consistent transcription • Better word understanding, especially in everyday speech • A natural conversation flow that doesn’t break when the user pauses or re-phrases Primary use-cases The pathway has to handle casual conversations and customer service inquiries without stumbling. Think quick back-and-forth chats where clarity and tone matter. Acceptance criteria • A/B voice tes...

Our client runs a call center with Nigerian agents calling US businesses. They need a Python script that makes the agents' accent sound more neutral and clear to American ears in real time during live phone calls. The hard part is already done. There is a fully built Node.js server that handles the live call audio. We just need you to write one Python script that plugs into it. What the script needs to do: Receive live audio from the server, process it through a voice conversion model to soften the Nigerian accent, and send the processed audio back. The whole process must happen in under 150ms so the conversation feels natural with no delay. Tech details: Audio comes in as mulaw 8kHz chunks. You can resample internally for better model performance then convert back before sending ...

I need a turnkey solution that converts my voice in real time on my MacBook Air M2 (16 GB, Apple Silicon). The cloned output must preserve tone, naturalness and emotional nuance while keeping latency to an absolute minimum—that is the single most critical factor for me. The system must run natively on macOS; Windows-only toolchains are not an option. I would like it wired into my daily workflow, specifically so it can be used during real-time voice calls (e.g., WhatsApp). Whether you rely on RVC, so-vits-svc, or another comparable model is up to you, as long as performance and stability match the spec. Before we start, I’ll need: • a clear, plain-English walkthrough of the architecture and how it will operate on Apple Silicon • a live session where I can hea...

Multilingual Voice Recording Project – Code-Switching Conversations Project Name: BV Project Type: Remote | Ongoing (Limited Slots per Locale) Project Overview Project BV is a multilingual speech data collection initiative designed to enhance Automatic Speech Recognition (ASR) systems for high-value multilingual call center scenarios, including financial services, healthcare, and telecommunications. The project focuses on collecting natural code-switching conversational audio, where speakers alternate between two languages within a single conversational turn. Scripts will be provided; however, natural delivery, fluency, and context-appropriate language switching are essential. Language Requirements Primary Language (Native level – one required): Catalan (ca-ES: ...