Automated Transcripts
Generate a verbatim text record of spoken content that can be reviewed and downloaded as part of the accessibility workflow.
Featured Case Study
Artificial Intelligence
completed
accessify-video
An AI-assisted video accessibility and remediation platform that transforms source video into transcripts, captions, audio descriptions, synthesized voice tracks, and remediated video through an automated processing workflow.
accessifyvideo.com
SSL
Project Information & Key Details
- Role
- Full-Stack Software Engineer
- Client Type
- Internal Product
- Industry
- Artificial Intelligence
- Duration
- 3 Months
- Team Size
- 2 Developers
Case Study Overview
Making video accessibility scalable
Captions make spoken content available as text, but they do not communicate important visual information to blind and low-vision viewers.
Creating accessible video can require several separate workflows: transcription, captioning, audio-description writing, voice production, timing, media editing, and final delivery.
AccessifyVideo was built to bring those stages into a single application and processing pipeline, reducing the manual work required to move from a source video to usable accessibility deliverables.
Case Study Overview
One video in. Multiple accessible deliverables out.
AccessifyVideo turns a single source video into a coordinated set of accessibility deliverables.
The processing workflow can generate a verbatim transcript, timed captions, an audio-description script, synthesized description audio, and a remediated video with the description voice incorporated into the original media.
Because these outputs originate from the same source file, the platform can coordinate captions, descriptions, timing, and remediation within one workflow rather than treating them as disconnected services.
For cases that require human review, the Editing Lab provides cue-level control over description wording and timing before re-rendering.
Platform Features
Timed Captions
Produce timed caption files from the source media for use with compatible video players, learning platforms, and content systems.
Audio Description
Analyze meaningful visual information and generate description cues designed to fit available spaces around dialogue.
Description Voice
Convert individual description cues into synthesized speech so edited cues can be regenerated without rebuilding the entire voice track.
Video Remediation
Use FFmpeg-based media processing to combine original audio, description voice, timing adjustments, and extended captions into a remediated video.
Editing Lab
Review and refine description cues at timeline level by editing wording, adjusting timing, adding or removing cues, and re-rendering the affected output.
Technical Architecture
Architecture designed around long-running media jobs
AccessifyVideo combines a React and TypeScript web application with a Node.js API, PostgreSQL-backed application state and job processing, AI-assisted media analysis, neural text-to-speech, and FFmpeg-based remediation.
The architecture deliberately keeps durable processing state in PostgreSQL so orders, jobs, retries, and progress can be coordinated without introducing a separate queue datastore.
Engineering Workflow
Building a resilient media-processing workflow
Video processing behaves differently from a typical web request. Individual jobs can run for long periods, consume significant disk and compute resources, and must remain recoverable when a worker or deployment is interrupted.
The implementation therefore treats media generation as a durable workflow rather than a single request-response operation.
Performance & Impact
Measurable Engineering Outcomes
- 6
- Accessibility deliverables from one source workflow
- 4
- Parallel analysis stages in the core media pipeline
- 1
- Source video coordinated across the generation workflow
- 1
- Maximum completed work that needs to be repeated after an interrupted processing run.
Case Study Overview
Technical Highlights
The platform combines AI, media processing, cloud infrastructure, and resilient backend architecture. It supports resumable rendering, secure authentication, payment processing, and scalable deployment for production workloads.
FAQ
Project questions
Building a complex digital product?
I build full-stack applications that combine user-facing experiences with reliable backend workflows, integrations, and production-ready engineering.
Portfolio Showcase
Explore More Case Studies
Live
Edtech
eQuraneKareem
A comprehensive online Quran education platform designed to connect students worldwide with Quran, Arabic, and Islamic courses through structured course discovery, tutor information, educational resources, and online enrollment.
Live
Edtech
CanadianGrad Academy
A full-stack online education platform for CanadianGrad Academy, combining course discovery, enrollment journeys, educational content, academic planning tools, and a custom administration system.