Spectromood

Spectromood: Speech emotion detection tool that uses a vision model to judge speaker feelings from voice measurements.

SpectroMood is an open experiment testing whether a general-purpose decision model can recognise emotion in speech using spectrogram plots and voice measurements. It checks audio windows for voice activity using Silero VAD, measures pitch, loudness, spectral tilt, harmonics-to-noise ratio, and speech rate, and compares them against a calm speaker baseline. The tool generates three stacked visual panels showing spectrogram, pitch contour, and loudness envelope. It sends these images alongside a typed question and a guide to Cloudflare's clef-flash model, which returns emotion probabilities and an arousal score. The software runs in a Docker container and supports English, Russian, and Spanish datasets.

Spectromood at a glance

Best for: Software engineer, Researcher
Runs on: Web, self-hosted, open source
Pricing: Free on Cloudflare Workers Free plan; API usage costs roughly $0.09 per million input tokens.
Released: October 2026
Category: AI research tool
Website: github.com/ketul93/spectromood

Key features

Silero voice activity detection gate
pYIN pitch tracking and loudness measurement
Spectrogram and pitch contour generation
Speaker calm baseline comparison
Cloudflare Clef decision model integration
Microphone and audio file analysis
Docker container setup

Use it for

Detecting speaker emotions from microphone input
Analyzing audio files for emotion and arousal
Running speech emotion benchmarks

Alternatives to Spectromood

Speaches — It acts as an Ollama equivalent for running local TTS and STT models.
SenseVoice — It runs over five times faster than Whisper-Small with low-latency inference.
Homi — It answers complex audio questions natively with high accuracy.
Overview — Learn language models by running hands-on experiments without writing scripts
MTEB — Evaluates embedding models across over 1000 tasks and 1000 languages.
LMMs-Eval — Provides open research and tooling for multimodal model evaluation and benchmarks.
Decido — Runs locally on CPU while beating published benchmark scores.
Hemory — Connects real-world conversations to your AI agents via MCP.
Desert Ant — Run fast on-device models with zero token costs and no cloud bills.

All Spectromood alternatives