Rev AI

Submits audio files via URL-based source configuration to a job queue that processes transcription asynchronously, returning job metadata with status tracking. Clients poll the job endpoint to retrieve transcript JSON containing monologues with speaker labels, word-level timestamps, and forced alignment precision. Built on 7M+ hours of human-verified speech data with proprietary ASR model optimized for conversational and telephony audio across 57+ languages.

Processes audio streams in real-time, delivering transcription results with minimal latency for live conversation, telephony, and broadcast scenarios. Streaming endpoint architecture enables continuous audio ingestion with incremental transcript updates, supporting speaker diarization and custom vocabulary injection during active sessions.

Returns transcription results in a structured JSON format with monologues array containing speaker-attributed segments, each with elements array containing individual words with type, value, start timestamp (ts), and end timestamp (end_ts). Custom media type application/vnd.rev.transcript.v1.0+json indicates structured transcript format with versioning, enabling backward compatibility and future schema evolution.

Accepts audio files for transcription via HTTPS URLs in the source_config object rather than direct file upload, enabling transcription of remote audio without client-side file transfer. URL-based submission reduces bandwidth requirements and enables transcription of large files, streaming sources, and cloud-stored audio without downloading to client machines.

Provides SOC II Type II, HIPAA, GDPR, and PCI DSS compliance certifications with 99.99% uptime SLA, encryption at rest and in transit, and dedicated HIPAA-compliant deployment options. Compliance infrastructure enables use in regulated industries (healthcare, finance, legal) with documented security controls and audit trails.

Provides Model Context Protocol (MCP) server implementation enabling integration with AI-powered code editors (Cursor, VS Code with MCP extension) for direct transcription access within editor environments. MCP server exposes Rev AI transcription capabilities as tools available to AI assistants, enabling in-editor transcription workflows without context switching.

Automatically identifies and labels distinct speakers in multi-party audio, attributing transcript segments to individual speakers with numeric speaker IDs. Diarization output is embedded in transcript JSON monologues structure, enabling downstream analysis of conversation patterns, turn-taking, and speaker-specific metrics without separate speaker identification API calls.

Injects domain-specific terminology, proper nouns, and technical jargon into the ASR model during transcription to improve recognition accuracy for specialized vocabulary. Custom vocabulary is submitted as a list and applied to both asynchronous and streaming transcription jobs, enabling accurate transcription of industry-specific terms, product names, and technical concepts without model retraining.

Generates precise word-level timestamps for each transcribed word, enabling frame-accurate synchronization between transcript and audio. Forced alignment API aligns transcript words to audio frames using dynamic time warping or similar alignment algorithms, producing start and end timestamps (ts, end_ts fields) for each word element in the transcript JSON output.

Automatically identifies and extracts key topics, themes, and subject matter from transcribed audio content using NLP analysis on the transcript text. Topic extraction API analyzes monologues and segments to surface primary topics discussed, enabling content categorization, search indexing, and conversation summarization without manual review.

Analyzes emotional tone and sentiment expressed in transcribed audio, classifying speaker sentiment as positive, negative, or neutral at the monologue or segment level. Sentiment analysis API processes transcript content and optionally audio prosody features to determine emotional valence, enabling conversation quality scoring, customer satisfaction assessment, and agent performance evaluation.

Automatically detects the language spoken in audio and routes transcription to the appropriate language-specific ASR model from 57+ supported languages. Language identification operates at the beginning of transcription jobs, with optional explicit language specification for improved accuracy. Supports multilingual audio with language switching detection within a single recording.

Delivers asynchronous notifications to a specified webhook URL when transcription jobs complete, eliminating the need for continuous polling of job status endpoints. Webhook system sends HTTP POST requests to client-specified endpoints with job completion metadata and transcript availability, enabling event-driven architectures and reducing API call overhead in production systems.

Implements OAuth-style Bearer token authentication for API access, with tokens generated and managed through the Rev AI web dashboard. Tokens are displayed once upon creation and must be securely stored by clients; maximum 2 active tokens per account enable key rotation and multi-environment deployments without account credential sharing.