Show HN: SpectroMood, speech emotion detection with a decision model

Category: ai-ml

Tags: speech-emotion-recognition, cloudflare-workers-ai, multimodal-llm, audio-analysis, benchmark

Score: 6.3/10 (Innovation: 6, Technical: 6, Documentation: 8, Utility: 5)

SpectroMood explores whether a general-purpose decision model (Cloudflare Clef) can recognize speech emotion from labeled plots and relative acoustic features, without audio-specific training. It includes a full pipeline (VAD, feature extraction, plotting, prompting) and a benchmark across English, Russian, and Spanish datasets, reporting modest but above-chance accuracy. The honest reporting of what works and what doesn't makes it a valuable experiment, though practical utility is limited.

Target audience: AI researchers, ML engineers, and developers exploring multimodal model capabilities

Repository: https://github.com/ketul93/spectromood · Python · Apache-2.0 · 3 stars

View on Hacker News