Ediphi Logo

Ediphi AI Interview App — Avatar Architecture & Cost Analysis

Technical evaluation: Cloud-Rendered Video (Azure) vs. Open-Source 3D (WebGL) vs. Modern Sound Orbs (Gemini / Grok / Siri).

3D Model / WebGL Digital Human
Option 1 · 3D Mesh

TalkingHead.js (WebGL)

Three.js Open-Source 3D Engine
  • Real-Time Lip-Sync: Browser Web Audio analyzes voice and automatically moves 3D mouth blendshapes (visemes: A, E, I, O, U) on student GPU.
  • Natural Motion: Real-time eye blinks, breathing animations, and subtle head tilts in browser.
  • $0.00 Video Compute: Renders directly on student hardware with zero cloud video bills.
2D Uniform Characters (pixi-live2d)
Option 2 · 2D Illustrated

pixi-live2d-display

Hospitality & Flight Attendant Tracks
  • Tailored Uniform Tracks: Hotel concierge, front desk blazer, and flight attendant industry attire.
  • Ultra-Lightweight (< 2 MB): Smooth volume-driven mouth flaps and hair physics on low-end school tablets.
  • $0.00 Video Compute: Pure client-side WebGL canvas rendering.
RenderPeople 3D Scanned Human
Option 3 · Scanned 3D Human

RenderPeople Scans

Photogrammetry Real People (renderpeople.com)
  • Can RenderPeople speak? Yes! Choose models from their "Rigged People" line with included facial rigs.
  • Viseme Lip-Sync: Facial blendshapes map to speech visemes for real-time speech via TalkingHead.js.
  • One-Time Purchase (~$40–$70): Permanent asset ownership with zero monthly video streaming costs.
Option 4 · Recommended

Fluid Sound Orb

Gemini Live / Siri / Grok Pattern
  • Modern Big Tech Standard: Minimalist glowing aura used by Google Gemini Live, Apple Siri, and OpenAI.
  • Zero "Uncanny Valley": Clean and high-tech—eliminates awkward robotic facial desync or lag.
  • Universal 60 FPS: Flawless performance on 100% of devices, even over 3G school Wi-Fi.
TEST LIVE ORB STATES:
🔄 100% Shared Backend: All 4 Options Run Client-Side with Just 2¢ / Interview

All visual options render locally on student devices with zero cloud video lag and connect to the exact same ultra-fast Gemini voice stream. Students can freely switch between 3D Characters, 2D Uniforms, Scanned Humans, or the Sound Orb with a single click!

$0.00 Video Bill

Executive Cost & Architecture Comparison

Per-interview unit economics: Microsoft Azure Legacy Cloud Video vs. Ediphi 3D Model / WebGL (Open Source Models) & Fluid Sound Orb.

1. Operating Cost per Completed Student Interview
Based on standard 4-minute session (3–6 questions) · Full stack: Video/Visual + Speech + AI Intelligence
Legacy Video Architecture
Microsoft Azure Realistic Video Avatar
What is Azure's Realistic Avatar? Microsoft runs deep learning models on cloud GPUs that generate photorealistic video frames of a real human person speaking, streamed to the student via WebRTC HD video. Because Microsoft must synthesize and encode full HD video every single second on cloud GPUs, it incurs a crushing $0.50/min ($30.00/hr) video rendering bill.
$2.16 / completed interview
~$0.540 / active minute ($32.40 / hr)
  • ✕ $2.00 / interview ($0.50/min): Cloud GPU video frame synthesis generates full HD video continuously
  • ✕ $0.16 / interview: Separate fragmented billing for Azure Speech-to-Text + OpenAI GPT-4o + Neural TTS
  • ✕ 3,000–5,000 kbps HD stream: Demands heavy constant video bandwidth; classroom computer labs buffer or fail on Philippine school Wi-Fi
  • ✕ 1.5s – 3.0s latency: Video encoding delays create awkward pauses and robotic conversation disconnects
❌ Total Cost a Month: $1,080 (at 500 interviews · $2.16 / session)
Recommended Architecture
Ediphi Next-Gen: 3D / 2D Mesh or Fluid Sound Orb
Zero Cloud Video Compute: Visuals render 100% locally on the student's phone or laptop using free WebGL/Canvas (TalkingHead.js, pixi-live2d, RenderPeople, or Fluid Sound Orb) while streaming only lightweight 32 kbps Gemini voice audio.
$0.02 / completed interview (2 Cents!)
~$0.005 / active minute ($0.30 / hr) (99.1% Cheaper)
  • ✓✓ $0.00 / interview: $0 video rendering fees (runs entirely on client GPU/Canvas)
  • ✓✓ $0.02 / interview: Unified native Gemini 2.0 Flash multimodal voice (4 min avg)
  • ✓✓ 32 kbps audio-only stream: 100x lighter than video; flawless on weak school Wi-Fi & 3G
  • ✓✓ Student choice: Free toggle between 3D Mesh, 2D Uniforms, Scanned Humans, or Fluid Orb!
🚀 Total Cost a Month: $10.00 (at 500 interviews · 2¢ / session)
2. Per-Interview Cost by Duration (3–5 Min Reality)
Direct comparison based on actual student interview sessions (3 to 6 questions each)
Interview Type Length Microsoft Azure Video Ediphi (3D Model / Sound Orb)
Quick Screening
3–4 introductory questions
 3 Mins $1.62 $0.015 (1.5¢) Save $1.60 (99%)
Standard Mock Interview AVERAGE
5–6 situational & behavioral questions
 4 Mins $2.16 $0.020 (2.0¢) Save $2.14 (99%)
In-Depth Mock Interview
7–8 hospitality competency questions
 5 Mins $2.70 $0.025 (2.5¢) Save $2.67 (99%)
Comprehensive Exam
10–12 questions + competency grading
 10 Mins $5.40 $0.050 (5.0¢) Save $5.35 (99%)
3. Architectural Reason (Why Ediphi is 99% Cheaper)
Cloud GPU video generation vs. Free client-side WebGL / Canvas rendering
Architecture Layer Microsoft Azure Video Avatar Ediphi (3D Model / Sound Orb)
Visual / Video Rendering
Cloud GPU vs. Local Device GPU
 $0.500 / min
$30/hr cloud GPU video encoding
 $0.000 / min
Free local WebGL 3D model or Canvas
Speech & AI Intelligence
STT + LLM Logic + Voice Audio
 $0.040 / min
Separate Azure STT + GPT-4o + TTS
 $0.005 / min
Unified Gemini Multimodal Live Audio
TOTAL RUNTIME / MINUTE ~$0.540 / min ~$0.005 / min (108x cheaper)
4. Real-World Monthly Volume Simulator: Ediphi Live Scale
Current production baseline: ~500 completed mock interviews / month across 3,000 enrolled students
Ediphi Live Production Scale
3,000 Enrolled Students · ~500 Monthly Interviews

In actual university and vocational tracks, students interview during checkpoint evaluations or job-placement season. With ~500 completed interviews monthly (~4 minutes each):

Total Cost a Month (Ediphi)
$10.00 / mo
Azure Video: $1,080 / mo
Quick Volume Presets:
Current load: ~500 interviews across 3,000 enrolled students
Standard 3–5 minute session (3 to 6 questions)
Unit Economics per Interview: $0.02 (2.0¢) on Ediphi vs. $2.16 on Azure
Total Monthly Traffic: 2,000 mins/mo (500 interviews @ 4 min)
Total Cost a Month (Azure Realistic Video)
$1,080 / mo
$2.16 per interview
Total Cost a Month (Ediphi Next-Gen)
$10.00 / mo
$0.02 (2.0¢) per interview
Prepared for Technical Review · Ediphi Training Programs Inc. · September 2026
// ========================================================= // GEMINI LIVE FLUID PLASMA ORB CONTROLLER // ========================================================= window.setOrbState = function(state, btn) { document.querySelectorAll('.mini-state-btn').forEach(b => b.classList.remove('active')); if (btn) { btn.classList.add('active'); } else { document.querySelectorAll('.mini-state-btn').forEach(b => { if (b.textContent.toLowerCase().includes(state.toLowerCase())) b.classList.add('active'); }); } const stage = document.getElementById('geminiStage'); if (stage) { stage.className = 'gemini-sound-stage ' + state; } }; // PER-INTERVIEW VOLUME SIMULATOR CONTROLLER function setPreset(interviews, btn) { document.getElementById('rngInterviews').value = interviews; document.querySelectorAll('.preset-btn').forEach(b => b.classList.remove('active')); if (btn) btn.classList.add('active'); recalculate(); } function recalculate() { const interviews = parseInt(document.getElementById('rngInterviews').value); const mins = parseInt(document.getElementById('rngMins').value); const totalMinutes = interviews * mins; // Unit costs per interview const azurePerInterview = mins * 0.54; const ediphiPerInterview = mins * 0.005; // Monthly totals const azureTotal = interviews * azurePerInterview; const ediphiTotal = interviews * ediphiPerInterview; // Update labels document.getElementById('lblInterviews').textContent = interviews.toLocaleString(); document.getElementById('lblMins').textContent = mins + ' mins'; document.getElementById('lblTotalMinutes').textContent = totalMinutes.toLocaleString() + ' mins/mo (' + interviews.toLocaleString() + ' interviews @ ' + mins + ' min)'; document.getElementById('lblUnitEdiphi').textContent = '$' + ediphiPerInterview.toFixed(2) + ' (' + (ediphiPerInterview * 100).toFixed(1) + '¢) on Ediphi'; document.getElementById('lblUnitAzure').textContent = '$' + azurePerInterview.toFixed(2) + ' on Azure'; // Update result cards document.getElementById('valAzure').textContent = '$' + Math.round(azureTotal).toLocaleString() + ' / mo'; document.getElementById('subAzure').textContent = `$${azurePerInterview.toFixed(2)} per interview`; document.getElementById('valEdiphi').textContent = '$' + (Math.round(ediphiTotal * 100) / 100).toFixed(2) + ' / mo'; document.getElementById('subEdiphi').textContent = `$${ediphiPerInterview.toFixed(2)} (${(ediphiPerInterview * 100).toFixed(1)}¢) per interview`; } recalculate();