Deepgram Nova-3 + Gemini Dual-Pass Audio Caption Studio

Deepgram API Key (Nova-3 STT Baseline)

Gemini Keys (Commas/Newlines for Shuffle & Failover)

⚡ Single 1-Upload Cloud Pipeline: Audio is uploaded ONCE to Cloudflare Pages in-memory proxy, which dispatches to Deepgram Nova-3 (with filler word detection) and Gemini in parallel over high-speed datacenter fiber for exact acoustic pronunciation sync.

Pass 2 Gemini Refiner

Output Script / Format

Spoken Language

Highlight Names & Sudden Expressions (e.g., "Zara", "Shrihari", "Aiyo!", "Wow!") Smart Emojis (Relevant Emotions & Key Nouns Only)

Select Audio File (MP3, WAV, M4A, WebM)

0%

Live Synchronized Captions (60 FPS Engine)

Process audio to test zero-lag synchronized captions.

100% Precision Word Timestamps (Click row to seek audio)

#	Word	Start (ms)	End (ms)	Duration (ms)	Type / Flags	Formatted Time
No refined data generated yet.

Full Refined Transcript

Deepgram Nova-3 Raw JSON Output (Pass 1 Baseline)

// Deepgram JSON will appear here...

Gemini Raw JSON Output

// Gemini JSON will appear here...