deepgram-stt

A real-time, AI-powered interview assistant that connects with platforms like Google Meet and Zoom to transcribe live interviews using Deepgram’s SST API (~100ms latency). It extracts questions, retrieves context from user-uploaded resumes or documents (via Gemini + Pinecone), and generates accurate responses using Claude.

Code snippets showing how to record I2S audio and store as .wav file on ESP32 with SD card, how to transcribe pre-recorded audio via Deepgram SpeechToText (STT) API, how to generate audio from text via TextToSpeech (TTS) API from OpenAI a/o SpeechGen a/o Google TTS. Triggering ESP32 actions via Voice.

AI-powered voice agent that joins meetings, transcribes speech in real-time using Deepgram STT, and responds intelligently.

Generate YouTube Shorts Video From an Article Link using Pydantic AI

AI-powered voice assistant for file and folder management. Built with LiveKit, OpenAI Realtime API, Deepgram API and Mem0 API, featuring rollback capabilities, audit logging, and comprehensive safety checks.

End-to-end AI interview platform featuring live voice interaction, coding assessment, and automated feedback.

EFL pronunciation assessment web/A2A app focused on syllable stress

A web-based telephony application bridging WebRTC and the PSTN (mobile & landline networks). Built with React and FastAPI, this system allows you to dial real phone numbers directly from your browser- no SIM card required. Powered by LiveKit SIP routing, Twilio PSTN trunking, and Deepgram real-time transcription

AI-native macOS menu bar dictation with a post-recording text pipeline.

Portfolio featuring real-time bilingual voice orchestration (Friday), agentic systems architecture, and live engineering telemetry.

Context aware ai smart glasses. (voice enabled)

AI-powered speech-to-text and text-to-speech platform with intelligent analysis using Deepgram. Features sentiment analysis, topic detection, intent recognition, and multiple voice personas.

Real-time voicemail detection using speech-to-text, beep frequency analysis, and silence monitoring for compliant automated message drops with prioritized decision logic.

Voice AI collections agent with a deterministic compliance layer and a standalone eval harness — built with Next.js, Groq (Llama 3.3 70B), and Deepgram. Verifies borrowers, negotiates repayment, and hands off to a human when it should.

A real-time audio transcription bridge that leverages WebSockets and Deepgram to provide low-latency, high-accuracy speech-to-text services.

AI驱动的面试辅助工具，实时捕获系统音频，自动转录考官提问并生成流式口语回答，适配国内AI模型，窗口隐身防屏幕分享检测。

Demo workflow for transforming French audio into AI-narrated English explanations.

Voice-first field survey app for UK ecologists. Talk while you walk a site; speech becomes BNG/PEA report fields, with completeness check flags. Next.js PWA + Deepgram + Claude.