Updated October 2026
How EchoDepth works
EchoDepth by Cavefish turns video input into decision signals in six steps. It detects 44 facial Action Units using the FACS standard, calibrates them against a cultural cohort and returns outputs such as a Trust Score and Credibility Rating, delivered by dashboard or report.
Input
EchoDepth receives video input - live feed, recorded session, or uploaded content - from the deployment environment.
Action Unit Detection
EchoDepth detects and classifies 44 facial Action Units using FACS-standard computer vision models.
Cultural Calibration
AU patterns are calibrated against the relevant cultural cohort from EchoDepth's multi-cohort reference set across multiple countries.
Communication Depth Mapping
Calibrated AU patterns are mapped across multiple delivery dimensions, producing a multi-dimensional communication quality output.
Signal Generation
Emotional state data is processed into decision-relevant signals: Trust Score, Credibility Rating, Resistance Indicator, Confidence Score.
Output Delivery
Outputs are delivered via dashboard or structured report (API delivery is planned) - formatted for the specific deployment context and governance requirements.
Real-time processing of all six steps, from raw video input to actionable decision signal, is planned and not yet available.
Every analysis passes through the same six steps, in order, from input to delivered output.
Sample EchoDepth reports
Structured, decision-ready output - formatted for executives, analysts and operational teams.
How does EchoDepth turn a video into a Trust Score?
EchoDepth takes video input from a live feed, recorded session or uploaded content and detects 44 facial Action Units using FACS-standard computer vision models, tracking delivery patterns at 25+ frames per second. Those patterns are calibrated against the relevant cultural cohort from EchoDepth's multi-cohort reference set across multiple countries. The calibrated patterns are then mapped across multiple delivery dimensions, producing a multi-dimensional communication quality output. For audio, the analysis also covers vocal pitch variation, speech pace, hesitation patterns and stress markers, and for text, linguistic tone and credibility mapping. All modalities combine into the composite Trust Score. Emotional state data is processed into decision-relevant signals, including the Trust Score, Credibility Rating, Resistance Indicator and Confidence Score. Outputs are delivered by dashboard or structured report and formatted for the specific deployment context and governance requirements. Sample analyses of existing content are delivered within 5 working days of submission. Real-time analysis of live sessions is planned and not yet available.
Common questions
How does EchoDepth actually analyse communication signals?
EchoDepth maps 44 Action Units (AUs) from the FACS standard in video, tracking specific delivery patterns at 25+ frames per second. For audio, it analyses vocal pitch variation, speech pace, hesitation patterns and stress markers. For text, it applies linguistic tone and credibility mapping across sentence structure and word choice. All modalities combine into the composite Trust Score.
Does EchoDepth require specialist cameras or hardware?
No. EchoDepth works on any standard RGB camera - a laptop webcam, phone camera or standard video recording is sufficient. For audio analysis, standard microphone recording quality is sufficient. No wearables, biometric sensors or specialist equipment are required.
How long does an EchoDepth analysis take?
A sample analysis of existing content (video, audio, text or image) is delivered within 5 working days of submission. Pilot programme analyses are typically delivered within 2–3 working days. Real-time analysis of live sessions is planned and not yet available.