AI Typer V2 — v0.5.5
Voice dictation with single-pass multimodal cleanup. Gemini 3 Flash is the new default; Debian package available.
Tag
gemini
8 posts
How much bitrate does an audio-multimodal LLM actually need?
An eval across 12 audio-capable models on OpenRouter, 5 MP3 bitrates, 4 dictation samples. The default upload quality in most dictation apps is 2–4× overprovisioned — and the more interesting finding had nothing to do with bitrate.
Testing Gemini 3.1 Flash Lite's Audio Understanding With 49 Structured Prompts
I built a 49-prompt test suite to evaluate Gemini 3.1 Flash Lite's audio understanding capabilities across 13 categories — from accent detection to deception analysis. Here's what worked, what didn't, and why it matters.
One Prompt AI Book: can Gemini 2.5 write a book in a single prompt?
Testing whether Gemini 2.5's 65K token output limit can produce a full book from one prompt. Spoiler: Anthropic did it better.
Using Gemini's vision capabilities for body language analysis
I built a body language analysis app using Google AI Studio's vibe coding interface and Gemini's multimodal vision. Upload a photo, get expert-level analysis.
Policy Visualiser: exploring global policy approaches with AI-powered clustering
An AI-powered React app that analyzes how different countries approach policy challenges, with interactive clustering visualizations powered by Gemini.
Voice Blog Creator: turning voice recordings into polished blog posts with Gemini
An automated pipeline that converts raw voice recordings into polished blog posts using audio preprocessing, Gemini transcription, and AI-powered formatting.
Voice Analyzer: an AI-powered voice analysis tool built with Gemini
A voice analysis application built with Google AI Studio and the Gemini API, exploring multimodal AI capabilities for audio processing.