oruk

oruk: Speech API for transcripts, emotion, and style

oruk is a speech API designed to process prerecorded English audio files. Built by researchers, the platform provides transcription, multilabel emotion detection, and speaking-style classification to help developers build speech products with deeper acoustic context.

Key Features

• English Transcription: Transcribes prerecorded English audio files using models like the flagship oruk Resonance for high accuracy.
• Multilabel Emotion Detection: Scores 15 emotion labels calibrated on held-out audio, allowing a single clip to carry multiple emotions and tracking changes over time with segment-level labels.
• Speaking-Style Classification: Provides 16 multilabel speaking-style outputs to describe how something was said.
• Unified Analysis: Delivers transcripts, calibrated labels, time-local segments, and tagged text in a single API request.
• Multiple Audio Formats: Supports mono or stereo inputs using WAV, FLAC, MP3, M4A, OGG, or WebM containers.
• Tiered Models: Offers Resonance for unified analysis, Spectra 1 for efficient affect workloads, and a preview tier, Spectra 2.

Use Cases

• Contact Center & Call Analytics: Useful for voice call sentiment analysis, AI call analytics, and voice of customer analytics.
• Business Intelligence: Applied in sales call analysis and meeting analysis.
• Research & Specialized Detection: Supports qualitative research and sarcasm detection.

Pricing

• Billing Structure: Billed by the measured second with a one-second minimum.
• Trial: Offers a $50 trial credit with no credit card required.
• Model Rates: Unified analysis costs $0.0120 per audio minute on Resonance and $0.0090 per audio minute on Spectra 1.

Getting Started

Developers can access the API by creating a verified account, receiving organization approval, and generating a production API key in the developer portal. The v1 API uses bearer authentication and multipart file uploads.

Website: https://oruk.ai/

oruk provides developers with a versioned API to extract words, emotions, and speaking styles from audio, offering scalable models for various speech understanding workloads.