Accessibility engineering for real-world communication
Lantern Arc builds accessibility engineering tools for AI, software, communication systems, and connected experiences.
We focus on whether critical information and communication remain usable for Deaf and Hard-of-Hearing people in real use.
How Lantern Arc works
Products
Lantern Arc is the parent brand. DeafAccess is the flagship accessibility platform in development. DeafBench is the benchmarking and evaluation product.
DeafAccess is in development. It is being built to help teams detect, prove, prioritize, fix, and retest communication failures affecting Deaf and Hard-of-Hearing users.
DeafBench measures accuracy, critical-information preservation, latency, and regressions across speech and captioning systems.
DeafAccess capability network
Labs are specialized testing capabilities inside DeafAccess, organized around the ways information can be lost, changed, delayed, or made unusable.
30-capability roadmap: these Labs are specialized testing capabilities inside DeafAccess, not separate companies or standalone brands.
Speech, captions, agents, context, critical information, and regression evidence.
Benchmark speech, captioning, and AI behavior with reproducible evaluation.
Test caption quality, completeness, timing, and critical-information preservation.
Evaluate communication workflows across chat, meetings, calls, and messaging.
Test voice and multimodal agents for accessible interaction and recovery.
Measure whether names, numbers, codes, dates, amounts, and warnings survive intact.
Test whether context needed to understand or act is preserved across modalities.
Run realistic end-to-end accessibility scenarios instead of isolated checks.
Stress communication systems under background noise and difficult input conditions.
Evaluate caption controls, readability preferences, and user-level customization.
Measure what gets lost between audio, text, visual, and haptic modes.
Detect accessibility regressions between product, model, or configuration releases.
Check evaluation data integrity, provenance, reproducibility, and evidence quality.
Alerts, emergency communication, real-time text, recovery, and public-space access.
Test whether important alerts have usable visual, text, or haptic equivalents.
Evaluate urgent communication flows where delay or information loss can raise risk.
Test real-time text continuity, timing, interoperability, and failure recovery.
Measure whether a user can recover when captions, alerts, or communication features fail.
Evaluate announcements, queues, kiosks, signage, and shared environments.
Support structured human review where automated evidence is not enough.
Sound cues, haptics, hearing devices, broadcast audio, vehicles, and vendor integrations.
Identify sound-only information and verify meaningful non-audio alternatives.
Evaluate vibration and tactile signals for clarity, timing, and equivalent meaning.
Test connected-device workflows involving hearing aids and cochlear implant ecosystems.
Evaluate broadcast-audio discovery, connection, control, and accessible fallback paths.
Test vehicle alerts, infotainment, navigation, and connected-car communication for non-audio access.
Evaluate third-party integrations and dependencies that affect accessible outcomes.
User experience, media, sign language, interpreting, DeafBlind access, and gaming.
Test whether accessible controls and workflows remain understandable and operable end to end.
Evaluate captions, transcripts, player controls, live media, and recorded content.
Evaluate sign-language experiences, presentation, controls, and modality handoffs.
Test interpreter access workflows, handoffs, scheduling, visibility, and continuity.
Evaluate combined visual, auditory, tactile, and text access requirements.
Test gameplay communication, cues, captions, multiplayer access, and recovery.
Open research
DeafBench gives Lantern Arc a public technical layer for evaluating speech, captioning, and AI systems with reproducible methodology and clear limitations.
