Guides
Build with the speech API
Working examples for sending audio, reading results, and using emotion and speaking-style scores in your application.
For the endpoint reference, see the API documentation. To call the API from a coding agent, see the MCP server.
Migration planning
Plan your Hume EVI migration
Inventory your voice stack, verify a customer-controlled export, and test a replacement conversation before cutover. Includes voice rights and failure checks.
Read →Python · Local ASR
Run Orukeet locally
Use OpenWhispr, native Python or sherpa-onnx with a supplied recording. Includes actual outputs, verified artifact hashes and troubleshooting.
Read →Python · Local review tool
Build a local call-review panel
Replay timed passages beside the transcript and separate vocal-expression scores. Run a complete Python example with public audio and a saved response.
Read →Node.js · Realtime preview
Stream transcription and phrase emotion
Run a complete Node.js WebSocket example, read phrase emotion scores, and handle final transcripts and session usage.
Read →Python · Resonance
Speaker diarization in Python
Label speaker turns with Resonance, read timed transcripts and emotion, and connect the results to a recording-review workflow.
Read →Python
Detect emotion from audio in Python
Inspect a real saved response, analyze your recording, and read selected emotion scores, timed segments, and speaker turns.
Read →Datasets
Speech emotion recognition datasets compared
Choose among IEMOCAP, MELD, RAVDESS, CREMA-D, and MSP-Podcast without confusing acted, conversational, and in-the-wild evaluations.
Read →Evaluation
How accurate is speech emotion recognition?
Read accuracy, macro F1, UAR, calibration, confidence intervals, and cross-corpus results without falling for a single headline number.
Read →Concepts
Voice sentiment analysis vs emotion recognition
Understand the difference between transcript sentiment, acoustic emotion, speaking style, and application-level intent.
Read →Audio ML
Prosody features for speech emotion and sarcasm
A practical guide to pitch, energy, timing, spectral shape, voice quality, and why no single cue proves an emotion or intent.
Read →