Case study · Media, Broadcasting & Accessibility

An AI-orchestrated accessibility pipeline for captions and audio description.

Phonetik turns standard video into accessible, regulation-ready media, generating closed captions, accessibility metadata and audio description tracks automatically.

Discuss a similar build All case studies

Client: Phonetik
Industry: Media, Broadcasting & Accessibility
Delivery: AI-driven media processing platform
Capability: Generative AI

The challenge

What Phonetik was up against.

Broadcasters and streaming platforms face growing accessibility mandates, but captioning and audio description remain manual, slow and inconsistent at library scale.

Manual captioning and audio description
Long turnaround for large libraries
Inconsistent subtitle quality and metadata
Little support for speaker attribution and sound context
Poor scalability across languages

The solution

What we built.

Skugal engineered an event-driven, agent-based platform combining computer vision, speech intelligence, sound classification and generative AI to process video at scale.

Open-to-closed caption conversion

Frame-by-frame text detection with Rekognition, OCR correction, timeline alignment and automatic SRT generation.

Speech, speakers & sound

Whisper transcription, PyAnnote and WhisperX diarisation, and YAMNet classification of music, ambience and non-speech cues, merged into timestamped JSON.

Automated audio description

Detects silent moments, extracts frames, generates concise visual descriptions with GPT-4 and narrates them with TTS, merged back with FFmpeg at frame precision.

How it works

From input to outcome.

Video upload
Frame & audio extraction agent
Speech & OCR agents
Speaker & sound classification agents
Caption & metadata assembly
Audio description generation
Accessible video + SRT

Business impact

What changed.

Phonetik delivered a production-grade accessibility platform that gives broadcasters faster compliance, lower operating cost and broader audience reach.

Technology