SM
Sheby M
Loading portfolio experience
Marketing · AI Agent, API, Web App
PolyVoice API – Multilingual STT & TTS
A FastAPI voice-processing solution combining OpenAI Whisper transcription, Coqui TTS voice generation, and LLaMA-based sentiment controls for English, Hindi, and Chinese.
Stack and focus areas
PythonVoice AIAIFull StackOpenAI WhisperCoqui TTS v2LLaMAFastAPISpeech-to-TextText-to-SpeechSentiment AnalysisVoice ModulationEnglishHindiChineseAudio ProcessingFFmpegDOCX GenerationPDF GenerationDockerAPI AuthenticationSaaS
Project description
The client needed separate but connected speech services that could transcribe uploaded audio, generate spoken audio from text, support English, Hindi, and Chinese, and change vocal delivery according to sentiment labels supplied by another AI model.
I developed PolyVoice API using Python and FastAPI. The STT service accepts uploaded audio, performs preprocessing, detects the supported language, and uses OpenAI Whisper to generate a transcript that can be downloaded as Word or PDF. The TTS service accepts text and uses Coqui TTS v2 to generate downloadable audio. A LLaMA integration classifies or accepts sentiment labels such as excited, sad, happy, calm, and professional, which are translated into controlled voice parameters before synthesis.
