Large existing back catalog
Years of YouTube videos published across departments, many without compliant captions or any audio description at all.
Vendor Proposal
ADA / WCAG 2.1 AA
A practical ADA video compliance workflow built for university media teams
Designed for large YouTube back catalogs, ongoing video production, and the real review, documentation, and quality work required for compliance—not just captions.
Built by someone already familiar with Stony Brook's video environment, workflows, and stakeholder expectations.
Why this matters now
The April 2026 deadline is an operational challenge, not a document problem
The DOJ's updated Section 504 rule requires that web content—including video—meet WCAG 2.1 AA. For a university with years of published video content across departments, this isn't resolved by a policy update.
It requires a repeatable production process: intake, transcription, human review, audio description, upload, and documentation. At scale, across multiple departments, over time.
- Accurate, synchronized captions on all public video content
- Audio descriptions for videos with meaningful visual content
- Human review, not just automated output
- Audit trail and completion documentation per video
- Processes that handle back catalog and new content on an ongoing basis
- Central tracking so departments aren't managing this ad hoc
The actual operational problem
This is not a small project with a simple handoff
University video compliance at scale involves several overlapping challenges. Any realistic vendor proposal needs to account for all of them.
Ongoing monthly production
New videos continue to be published. Compliance can't be a one-time sprint—it requires a steady intake process that fits existing workflows.
Captions are only part of the work
Audio descriptions require separate review and production—especially for content with meaningful visual information that isn't conveyed by dialogue alone.
Automated output isn't enough
AI transcription tools produce drafts. Human review, correction, and approval is required before output is compliant. That step needs to be built into the process.
Documentation and audit trail
Compliance requires more than completed files. There needs to be a record of what was reviewed, by whom, and when—per video.
Large vendors aren't always the right fit
Enterprise accessibility vendors offer broad platforms, but they may not be positioned for the responsiveness and hands-on flexibility a university media operation needs.
Why this is a strong fit
I already understand the environment
This is not a cold vendor pitch. I do contract work in Stony Brook's video department and understand the workflows, the people, and the production context from the inside. That changes what I can offer.
Embedded, not external
Because I already work in Stony Brook's video environment, I understand the production context, naming conventions, stakeholder preferences, and platform constraints better than a vendor engaging from scratch.
Built for this specific task
I am actively building a workflow system designed around exactly this use case: university YouTube back catalogs, recurring intake, captions, audio description, QA, and documentation in a managed loop.
Process, not just labor
What I am offering is a managed workflow with automation for the repetitive parts and human review built in. This is a system that can handle volume without proportional cost increases.
Responsive and lean
Lower overhead than an enterprise vendor means more flexibility in how the engagement is structured—starting small, adjusting the process, and scaling up without organizational friction.
Cost structure that fits the work
Enterprise vendor pricing is built to support large teams and platform infrastructure. My pricing model is built around the actual work per video—more transparent and predictable for budgeting purposes.
Realistic about what I can deliver
I am not promising to be a full enterprise accessibility firm. I am well-positioned for a specific scope: YouTube video compliance workflow for a university media operation. That specificity is a strength.
How the workflow works
A managed process from intake to documentation
This is not a service where you send videos and receive caption files. It is a managed production loop—automated where appropriate, with human review built in at every quality checkpoint.
The output is not just completed files. It is a documented, trackable compliance record per video.
Automated
System handles first draft
Human review
Manual QA checkpoint
"Every video goes through both automated and human stages before it is marked complete."
Video intake
Videos are pulled from YouTube playlists, channel exports, or provided CSV lists. Each video is logged in the tracking system with its metadata, status, and priority.
AutomatedFirst-draft caption generation
AI transcription (OpenAI Whisper or equivalent) produces a time-coded caption file. This draft is stored and flagged for review—not delivered as a finished product.
AutomatedHuman caption review and correction
A human reviewer checks the draft against the video, corrects errors, fixes speaker labels, and approves the file. Accuracy rate and review notes are logged.
Human reviewAudio description assessment
The video is assessed to determine whether meaningful visual information exists that requires audio description. Many videos may not require it; the ones that do are flagged for production.
Human reviewAudio description production
For videos requiring it: a script is drafted (AI-assisted), reviewed for accuracy and timing, and approved. The described audio is produced and integrated. This stage is handled carefully because it carries the most quality risk.
Human reviewUpload and delivery
Approved caption files are uploaded directly to YouTube via API. Audio description tracks are delivered in the agreed format. Status is updated in the tracking system.
AutomatedCompletion documentation
Each video receives a completion record: what was done, when, who reviewed it, and what the output files are. This is the audit trail that supports compliance documentation.
Automated
Tools and infrastructure
Lean, modular, and purpose-built
The stack is chosen for reliability and low operating overhead—not to impress with complexity. Each tool has a specific job in the workflow.
Workflow orchestration. Manages intake, triggers processing steps, handles API calls, and coordinates the end-to-end pipeline without manual intervention for routine work.
First-draft transcription. Produces time-coded caption drafts from video audio. Accuracy varies by audio quality; human review always follows.
Pulls video metadata, handles caption file upload directly to YouTube, and confirms delivery. Eliminates manual upload for batches of videos.
A lightweight internal tool for human reviewers to access draft captions, log corrections, and mark videos as approved—keeping quality gates in the workflow rather than in email.
Modular—the specific tool or provider used for audio description drafts can be swapped based on quality benchmarks. The review and approval step is fixed regardless of the tool.
Central log of every video in the system: intake date, current status, review completion, delivery confirmation, and documentation output. Exportable for compliance reporting.
Used selectively for audio description script drafting and for reviewing complex transcription edge cases. Not used as a replacement for human judgment in quality-critical steps.
Self-hosted infrastructure where applicable. No critical workflow steps depend on third-party platforms that could create data retention or compliance concerns.
"Nothing here is speculative. These are tools I use in production work. The workflow is being built around this specific use case."
How the models compare
Large vendor vs. lean managed service
This isn't a claim that large accessibility vendors do bad work. It's an honest comparison of what each model is built for, so stakeholders can evaluate fit.
Large enterprise vendor
Onboarding
Formal procurement, contract cycles, multi-week setup
Responsiveness
Account management layer; requests routed through support
Flexibility
Platform-driven; adjustments require change orders
Local knowledge
No existing relationship with Stony Brook's environment
Cost structure
Includes platform overhead; per-minute pricing at scale
Pilot risk
Larger upfront commitment to begin
Lean managed service (this proposal)
Onboarding
Pilot batch can begin quickly; no extended setup phase required
Responsiveness
Direct point of contact; no account management layer
Flexibility
Workflow adjustments are possible based on actual Stony Brook needs
Local knowledge
Existing relationship with Stony Brook's video environment and workflows
Cost structure
Built around per-video work; lower overhead than a platform subscription
Pilot risk
Low-commitment pilot is the designed entry point
Engagement model
Start with a pilot batch. Validate before committing.
The right way to evaluate this proposal is to run a small pilot—not to evaluate a deck. A pilot lets Stony Brook see the actual workflow, output quality, and turnaround time before making a larger commitment.
Pilot batch
A defined set of videos—back catalog or recent production—processed through the full workflow. Output reviewed together. Process adjustments made.
Workflow validation
Review caption accuracy, audio description quality, delivery format, and documentation output. Confirm that the workflow fits Stony Brook's review and approval preferences.
Back catalog remediation
With a validated workflow, begin processing the existing back catalog at an agreed pace and priority order. Progress tracked and reported weekly.
Ongoing monthly intake
New videos produced by the department enter the workflow on a recurring schedule. Compliance is maintained going forward, not just achieved once.