What a Virtual Event Platform actually does
Matches attendees to each other and to sessions using embedding similarity, generates sponsor lead summaries from booth-chat transcripts, and produces post-event highlight reels — turning raw conference data into actionable revenue intelligence for organizers.
A white-label AI virtual event platform layers three AI capabilities on top of proven WebRTC + expo-hall infrastructure (vFairs or Airmeet). First, attendee-to-attendee matchmaking uses text-embedding-3-small to embed attendee profiles (job title, interests, goals) and surfaces the top-10 recommended connections per person before the event opens — at $0.00002/profile this is essentially free. Second, session recommendation uses Claude Sonnet 4.6 to reason over an attendee's profile and the full session catalog to produce a personalized agenda. Third, post-event sponsor value delivery: Deepgram Nova-3 transcribes booth-chat recordings, Sonnet 4.6 extracts lead-qualification signals (budget, timeline, decision role) from each conversation, and outputs a structured lead summary that sponsors actually act on.
The market signal driving demand in 2026 is sponsor accountability. Post-pandemic virtual event budgets were slashed because sponsors couldn't prove ROI from virtual booths. AI lead-scoring changes that equation: an agency that delivers a structured 'here are your 23 qualified leads with budget signals and next-step recommendations' report commands a $2,000–$5,000 premium per event over platforms that only hand over a badge-scan CSV. vFairs and Airmeet have the most mature reseller programs with real white-label tiers — the AI layer on top is where the agency differentiates, not the WebRTC infrastructure.
AI capabilities involved
Attendee-to-attendee networking matchmaking via profile embeddings
Personalized session recommendation per attendee
Real-time multi-language captions across all sessions
Sponsor lead-qualification extraction from booth conversations
Post-event highlight reel generation
Who uses this
- Conference-production agencies running 3–30 multi-track events/yr for association and trade-show clients
- Association management organizations (AMOs) that produce annual conferences for professional societies
- Trade-show producers who pivoted to hybrid events in 2022–2024 and now need an AI-powered networking differentiator
- Event-tech consultants who want to bundle a branded matchmaking product into their event management retainer
- Corporate events teams producing internal summits with 500–5,000 attendees who need executive-level engagement analytics
SaaS alternatives on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
vFairs
Conference agencies with 1–5 large (2,000+ attendee) events/yr where per-event pricing is acceptable and the 3D expo hall is a client-requested feature.
Demo on request
~$5,000+/event (Pro+)
Full WL on Pro+ (quote-based)
Pros
- +Most mature reseller program in this category — agencies get full brand removal and their own domain across attendee-facing pages.
- +Virtual expo halls with 3D booth rendering are a genuine differentiator over Zoom/Teams alternatives.
- +Dedicated customer success rep for reseller partners — reduces support burden on the agency.
- +Built-in attendee matchmaking (though less configurable than a custom embedding pipeline).
Cons
- −Per-event pricing at $5K+/event makes the economics volatile for agencies with irregular event calendars.
- −The '3D expo hall' is a differentiator in 2021 terms — in 2026 many enterprise clients see it as visual noise rather than value.
- −API documentation is gated behind a partner agreement — plan 2–3 weeks lead time to get API credentials.
- −AI capabilities are platform-standard; sponsors cannot get customized lead-scoring logic.
Airmeet
Agencies running 8–20 professional-society or B2B-conference events/yr where networking quality matters more than 3D expo aesthetics.
Free plan (50 attendees, limited sessions)
$1,000+/mo (Conference plan with WL)
WL on Conference plan and above
Pros
- +Monthly flat-rate pricing is more predictable for agencies running 10+ events/yr than per-event vFairs.
- +Social lounge feature (round tables for networking) is more natural for conference networking than expo booths.
- +White-label on Conference plan includes custom domain, email, and removal of Airmeet branding.
- +Better API documentation than vFairs — faster integration for custom AI layers.
Cons
- −3D expo hall feature is less mature than vFairs — exhibitor clients with booth-heavy trade shows may prefer vFairs.
- −Social lounge capacity caps at 50 per table — large networking events need multiple rooms managed manually.
- −Support response times on Conference plan can be slow outside US business hours.
- −Matchmaking feature uses LinkedIn import rather than custom profile fields — limits AI embedding input quality.
Hopin / RingCentral Events
Agencies whose clients are already in the RingCentral ecosystem (Teams users) and want a hybrid event platform with an existing technology relationship.
Free (100 attendees, limited sessions)
$667+/mo (Premium with WL)
WL on Premium
Pros
- +Most recognizable brand in post-pandemic virtual events — clients may already be familiar with the platform.
- +RingCentral backing provides infrastructure stability after Hopin's financial difficulties.
- +Breakout sessions, stages, and expo all in one platform at a lower price point than vFairs.
- +Integration with RingCentral Rooms for hybrid events (physical + virtual).
Cons
- −Hopin's history of rapid feature cuts and platform instability (multiple pivots 2022–2025) creates vendor-lock risk.
- −Premium WL at $667/mo is the cheapest real WL option but has attendee caps that add per-head fees at scale.
- −AI features are less developed than Airmeet's or vFairs' — no built-in matchmaking on WL tier.
- −RingCentral's enterprise-sales focus means support for SMB agencies is not a priority.
Bizzabo
Enterprise-event agencies serving Fortune 500 clients where robust Salesforce integration and KLIK badge scans justify the $15K+ entry point.
Demo only
$15,000+/yr (Growth)
Partial WL on Enterprise
Pros
- +Best-in-class event data and engagement analytics — sponsors get more actionable reports than on vFairs or Airmeet.
- +Strong CRM integrations (Salesforce, HubSpot) native — reduces post-event data wrangling.
- +KLIK badge hardware integration for in-person event lead capture.
- +Built-in mobile app with agenda and networking.
Cons
- −At $15K+/yr it's the most expensive option in this list — only defensible for agencies with 5+ large enterprise clients.
- −WL is partial: Bizzabo's domain and branding appear in some email templates even on Enterprise.
- −Complex onboarding requires Bizzabo's implementation team — 4–6 week setup before first event.
- −No per-event pricing — annual contract locks you in regardless of event cadence.
The AI stack
The virtual event AI stack has three distinct jobs at three distinct cost levels: pre-event matchmaking (cheap, runs once), live session captioning (per-minute streaming, moderate cost), and post-event sponsor intelligence (the highest-value output, moderate cost). Design the pipeline to parallelize post-event jobs.
Attendee profile embeddings and matchmaking
Converts attendee profiles into vector embeddings, computes similarity scores, and surfaces top-N matches per attendee before the event opens
text-embedding-3-small
$0.02/M tokensAll standard B2B conference matchmaking — the cost-quality tradeoff is strongly favorable
Voyage-3-lite
$0.06/M tokensPremium events where attendees explicitly paid for 'high-quality curated networking' — the $0.04 extra per profile is defensible
Our pick: text-embedding-3-small for all standard events — the $0.02 total cost per 2,000-attendee event makes Voyage-3-lite's quality improvement hard to justify. Run embeddings 24 hours before event open so matches are ready at launch.
Session personalization
Generates a personalized agenda for each attendee based on their profile, past session attendance, and poll responses
Claude Sonnet 4.6
$3/$15 per M tokensLarge conferences (500+ sessions) where session-description context is dense and persona-matching requires nuanced reasoning
GPT-5.4 mini
$0.75/$4.50 per M tokensStandard B2B conferences with 20–100 sessions — the dominant use case
Our pick: GPT-5.4 mini for events under 200 sessions. Sonnet 4.6 for large conferences with 200+ sessions or attendees with detailed profile data.
Live transcription and captions
Converts live session audio into real-time captions and post-event searchable transcripts
Deepgram Nova-3 (streaming)
$0.0077/min ($0.46/hr)All live sessions where captions are displayed to attendees
Deepgram Nova-3 (batch, post-session)
$0.0043/min batchPost-session transcript processing for sponsor lead extraction — don't use streaming for non-live processing
Our pick: Nova-3 streaming during live sessions for captions; Nova-3 batch on recordings for post-event sponsor intelligence. At a 2-day conference with 10 simultaneous tracks and 8 hours/day, streaming costs $0.46 × 10 × 16 = $73.60 — budget this explicitly.
Sponsor lead intelligence
Extracts structured lead-qualification signals from booth-chat transcripts — the highest-value AI output for sponsor retention
Claude Sonnet 4.6
$3/$15 per M tokensAll sponsor lead summaries — this is the primary paid value and quality matters
GPT-5.4
$2.50/$15 per M tokensAlternative to Sonnet 4.6 when per-event cost needs to be under a specific budget
Our pick: Sonnet 4.6 for all sponsor lead summaries — this is the upsell feature sponsors pay for. The $22/event cost is trivial against a $2,000–$5,000 sponsor ROI report package.
Highlight reel generation
Produces a 60–90 second post-event sizzle reel for social media and next-year promotion
Veo 3.1 Lite
$0.05/sec 720pAnimated or branded graphic reel segments between actual speaker clips
Gemini 3.5 Flash (transcript-based clip selection)
$1.50/$9 per M tokensAgencies with a video editor on staff who want AI to pre-select the best moments from 8 hours of content
Our pick: Gemini 3.5 Flash transcript-based clip selection is the right default — output the top-10 timestamps to your editor for a $0.50 cost versus a $3 AI-generated reel that looks synthetic. Use Veo 3.1 Lite only for intro/outro animated segments.
Reference architecture
The architecture splits cleanly into three phases: pre-event (embeddings + session personalization), live (captions + real-time matchmaking nudges), and post-event (sponsor intelligence + highlight selection). The hardest engineering challenge is pre-event embedding freshness — profiles update as attendees complete registration up to event-day, so the matching pipeline must re-run incrementally, not just once.
Agency admin creates event in the platform, connects to vFairs/Airmeet via API, and configures sponsor tiers and booth mappings
Next.js admin dashboard + Supabase events table + vFairs APIEvent configuration stored in Supabase with tenant_id isolation. Sponsor booths mapped to Supabase sponsor_booths table with webhook endpoints registered for chat events.
Attendees register via custom domain — profile data synced from vFairs/Airmeet registration webhook
Supabase Edge Function receiving vFairs registration webhookEach new registration triggers an embedding job: profile text (job title, company, interests, goals) is embedded via text-embedding-3-small and stored in Supabase pgvector attendees table. Top-10 cosine matches computed and stored in matches table.
24 hours before event open: personalized agendas generated for all registered attendees
Trigger.dev scheduled job → GPT-5.4 mini Edge FunctionBatches attendees in groups of 50. Each batch call passes attendee profile + full session catalog. Returns ranked session list stored in attendee_agendas table. Sent via Resend as pre-event 'Your personalized agenda' email.
Event opens: live captions stream from Deepgram Nova-3 to all active sessions
Deepgram WebSocket streams per session track + Supabase RealtimeEach session has a Deepgram streaming connection via a Fly.io worker. Caption chunks arrive every 2 seconds, stored in captions table keyed by session_id + timestamp. Broadcast via Supabase Realtime to caption overlay on session page.
Sponsor booth conversations are recorded and chunked via vFairs/Airmeet webhook
vFairs/Airmeet booth-chat webhook → Supabase booth_conversations tableEach booth conversation (video call + text chat) is stored with speaker metadata. At conversation end, Nova-3 batch transcribes any audio. Transcript stored alongside chat messages.
Post-event: Sonnet 4.6 extracts lead-qualification signals from each booth conversation
Trigger.dev post-event job → Sonnet 4.6 Edge FunctionEach conversation transcript gets a structured extraction: contact name, role (decision-maker / influencer / champion), budget signal (mentioned or inferred), timeline (immediate / 3-6 months / 12+ months), next-step recommended, overall qualification score 1-10. Output stored in booth_leads table.
Sponsor lead report published to sponsor portal — structured PDF + live dashboard
Next.js sponsor portal page + Supabase booth_leads + React PDFEach sponsor sees only their booth's data (RLS by sponsor_id). Lead table sortable by qualification score. PDF export via React PDF. Report is the primary deliverable for sponsor retention and upsell.
Post-event: Gemini 3.5 Flash selects top-10 highlight moments from session transcripts for video editor
Trigger.dev post-event job → Gemini 3.5 Flash Edge FunctionFull transcript corpus passed to Gemini 3.5 Flash with instructions to identify the 10 highest-energy/most-quotable moments. Output: [{session_title, timestamp, speaker, quote, reason}]. Stored in highlights table and emailed to event producer.
Estimated cost per request
~$4 per 2,000-attendee event in AI costs: $0.02 embeddings + $0.90 agenda gen (2,000 × $0.00045) + $73.60 captions (10 tracks × 8 hrs × $0.46) + $22 sponsor lead summaries (50 booths × 20 chats × $0.022) + $0.50 highlight selection. Captions dominate — budget per-track-hour, not per-attendee.
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
Models AI costs for a single multi-track virtual conference. The dominant cost driver is Deepgram streaming transcription across concurrent session tracks — not matchmaking or sponsor intelligence.
Estimated monthly cost
$150
≈ $1,806 per year
Calculator notes
- Deepgram streaming at $0.46/hr per track is the dominant cost — model this carefully. A 3-day, 8-track, 8-hr/day conference = $88 in transcription alone.
- vFairs or Airmeet platform fees ($1,000–$5,000+/event) are NOT included in this calculator.
- Sponsor lead summary cost assumes an average of 20 booth conversations per sponsor at ~$0.022 each via Sonnet 4.6.
- Attendee matching re-runs every 24 hours during registration period — multiply by number of days from registration open to event start for total embedding cost.
Build it yourself with vibe-coding tools
By Sunday you can have an AI matchmaking dashboard and sponsor-lead portal running on top of your vFairs or Airmeet account — without touching streaming infrastructure.
Time to MVP
12–16 hours (1 weekend)
Total cost to MVP
$25 Lovable Pro + $1,000+ vFairs/Airmeet (first month) + ~$40 LLM API credits
You'll need
Starter prompt
Build a white-label virtual event AI layer using Next.js App Router and Supabase. This app sits on top of vFairs or Airmeet and adds AI matchmaking, personalized agendas, and sponsor lead intelligence. Supabase schema: - events (id, tenant_id, name, vfairs_event_id, start_date, end_date, status) - attendees (id, event_id, email, full_name, job_title, company, interests_text, embedding vector(1536)) - sessions (id, event_id, title, description, track, start_time, speakers, embedding vector(1536)) - matches (id, attendee_id, matched_attendee_id, similarity_score, viewed) - sponsor_booths (id, event_id, company_name, tier) - booth_conversations (id, sponsor_booth_id, transcript_text, lead_name, lead_email, lead_role, budget_signal, timeline, qualification_score) - All tables RLS-isolated by tenant_id Pages to build: 1. /admin — event creation + attendee/session CSV import 2. /[event]/networking — attendee-facing matchmaking page showing their top-10 matches with shared interests highlighted 3. /[event]/my-agenda — personalized session agenda for each logged-in attendee 4. /[event]/sponsor/[booth_id] — sponsor portal showing their qualified lead table with filter/sort 5. /[event]/organizer — post-event dashboard: highlight moments, session engagement stats, sponsor summary Use Tailwind CSS + shadcn/ui. Auth via Supabase with magic link (no passwords for attendees).
Paste this into Lovable
Follow-up prompts (run in order)
- 1
Add a Supabase Edge Function that takes an attendee's profile text (job_title + company + interests_text), calls OpenAI text-embedding-3-small, stores the 1536-dim vector in attendees.embedding via pgvector, then queries the 10 most similar attendees using cosine distance. Store results in the matches table. Trigger this on every new attendee registration webhook from vFairs.
- 2
Add a Trigger.dev job that runs 24 hours before event start. For each attendee, call a Supabase Edge Function that passes their profile + the full session list to GPT-5.4 mini and returns a JSON array of top-8 recommended sessions with one-sentence explanations. Store in attendee_agendas table and trigger a Resend email with the personalized agenda.
- 3
Add a Supabase Edge Function for sponsor lead processing. Input: booth_conversation_id. Fetch the transcript. Call Sonnet 4.6 with a structured extraction prompt that returns: {contact_name, contact_email, role_type: 'decision-maker'|'influencer'|'champion'|'unknown', budget_mentioned: boolean, budget_range: string|null, timeline: '0-3mo'|'3-6mo'|'6-12mo'|'12+mo'|'unknown', qualification_score: 1-10, next_step: string, summary: string}. Store in booth_conversations. Trigger for each conversation when it ends.
- 4
Add a Deepgram streaming proxy Supabase Edge Function that accepts a session_id, opens a Deepgram Nova-3 WebSocket connection configured for the session's audio source, and writes caption chunks to the captions table every 2 seconds. The caption chunk includes {session_id, timestamp_ms, speaker, text}. Broadcast via Supabase Realtime so the attendee-facing session page can show live captions.
- 5
Add a post-event highlight reel selector: a Trigger.dev job that runs when event status changes to 'ended'. It fetches all session transcripts, calls Gemini 3.5 Flash with the prompt 'From these session transcripts, identify the 10 most memorable, quotable, or high-energy moments suitable for a post-event sizzle reel. Return [{session_title, timestamp, speaker, quote, reason}]'. Store in highlights table and email the organizer.
Expected output
A branded event portal where attendees see their top-10 networking matches and personalized agenda, sponsors get a live lead-qualification dashboard, and organizers have a post-event highlight selector — all powered by your AI pipeline and presented under your agency's brand.
Known gotchas
- !vFairs API access requires a signed partner agreement — start paperwork at least 2 weeks before your first client event. Building with the API without access is not possible.
- !pgvector cosine distance queries slow down dramatically above 10,000 attendees without an HNSW index — add CREATE INDEX ON attendees USING hnsw (embedding vector_cosine_ops) to the Supabase migration.
- !Lovable may generate the embedding storage as JSONB instead of the native pgvector type — manually fix the column type in the Supabase table editor before importing attendee data.
- !Deepgram streaming requires a server-side proxy (not a browser WebSocket) to avoid exposing the API key — Lovable's generated code often puts this on the client. Move it to a Supabase Edge Function.
- !Trigger.dev's free tier caps at 10,000 job runs/month — a 2,000-attendee event with 3 jobs per attendee (embedding, agenda, post-event) = 6,000 runs. Works on free tier for one event; upgrade for recurring events.
- !Sponsor booth conversations from vFairs include both text chat and video call metadata — video call transcription requires separate Deepgram batch processing after the recording is available, not real-time.
Compliance & risk reality check
Virtual events that transcribe sessions, profile attendees for AI matchmaking, and extract lead-qualification signals from booth conversations create meaningful compliance obligations across data privacy, AI transparency, and sponsor-data handling.
GDPR + recording consent for EU attendees
Recording EU attendees' voices in sessions or booth conversations is biometric data processing under GDPR Article 4. Sponsors receiving AI-extracted lead data from EU attendees who did not explicitly consent to their conversation being analyzed for commercial purposes face GDPR Article 6 lawful-basis challenges. Data processor agreements are required between the agency, the event organizer, and every sponsor.
Mitigation: Add explicit 'I consent to session recording and AI transcription for matchmaking and lead-generation purposes' checkbox during registration. Pre-populate session recording consent in the event platform. Include sponsor DPA templates in the client contract. Route EU session transcripts through Mistral Large 3 on EU-hosted inference.
EU AI Act Art. 50 — AI matchmaking and recommendation disclosure
From August 2, 2026, AI Act Article 50 requires disclosure when AI systems make recommendations about people to other people. The matchmaking feature (AI recommending attendee A to attendee B) and the session recommendation engine both fall under this disclosure requirement for EU event participants.
Mitigation: Add 'Connections and session recommendations are suggested by AI based on your profile' disclosure on the networking and agenda pages. Include in the event FAQ. No opt-out is required, but disclosure is mandatory.
Sponsor lead data: GDPR/CASL/CCPA opt-in for follow-up emails
AI-extracted lead data (name, company, buying signals) passed to sponsors constitutes a data transfer for marketing purposes. Under GDPR Article 6(1)(a), this requires explicit consent at collection — the attendee must know their booth conversation will be summarized and shared with the sponsor. Under CASL, the same logic applies for Canadian attendees. Under CCPA, California attendees have the right to opt out of sale/share.
Mitigation: Display 'Conversations in sponsor booths may be recorded, transcribed, and summarized for lead-qualification purposes' on the booth entry page. Add a 'no-record' toggle for attendees who opt out. Exclude opted-out attendees from sponsor lead reports. Include this disclosure in the event registration flow.
SOC 2 Type II for enterprise event organizers
Enterprise association or trade-show clients will require SOC 2 Type II before signing annual contracts — storing attendee PII, sponsor API credentials, and session transcripts in Supabase creates scope for the audit.
Mitigation: Use Supabase SOC 2 compliant infrastructure (Pro+). Encrypt all API keys in Supabase Vault. Store session transcripts with 90-day auto-deletion policies for events without explicit retention requirements. Run Vanta ($4K+/yr) for automated SOC 2 evidence collection.
WCAG 2.2 AA captions for public-sector events
ADA Title II final rule (effective April 24, 2026) requires WCAG 2.1 AA conformance for state and local government digital content. Virtual events produced for government or public-sector associations must provide real-time captions and ensure replay captions are synchronized and downloadable.
Mitigation: Use Deepgram Nova-3 streaming for live captions displayed overlaid on all sessions. Store timestamped VTT caption files alongside recordings in Cloudflare R2. Provide caption download link in the replay portal. Test caption accuracy on a 30-minute sample before each event.
Build vs buy: the real math
4–6 weeks for AI matchmaking layer on vFairs/Airmeet
Custom build time
$20,000–$40,000
One-time investment
4–6 months
Breakeven vs buying
At $20,000–$40,000 for the RapidDev AI layer on top of vFairs/Airmeet, the breakeven math is straightforward: if the AI sponsor-intelligence report adds a $2,000 premium to your event package and you run 12 events/year, you recover the build cost in 10 events ($20,000) to 20 events ($40,000) — i.e., 10–20 months. The faster-break-even case: if you convert 3 sponsor contracts from $5,000 to $8,000/event because you can promise structured lead qualification, the AI layer pays for itself in 7–13 events. As Sonnet 4.6 prices continue falling (expect 30% reduction by mid-2027 based on T8 decay curves), your AI COGS margin widens without changing the sponsor price — pure margin expansion.
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship.
Discovery call (free)
30 minWe map your exact Virtual Event Platform use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
4–6 weeks for AI matchmaking layer on vFairs/AirmeetOur engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
4–6 weeks for AI matchmaking layer on vFairs/Airmeet
Investment
$20,000–$40,000
vs SaaS
ROI in 4–6 months
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does it cost to build a white-label AI virtual event platform?
Two very different answers depending on what you're building. An AI matchmaking and sponsor-intelligence layer on top of vFairs or Airmeet (which handle all streaming infrastructure) costs $20,000–$40,000 with RapidDev and delivers in 4–6 weeks. Building the full WebRTC streaming, expo-hall, gamification, and sponsor analytics stack from scratch costs $150,000–$400,000 and takes 25–40 weeks. For 99% of conference agencies, the AI layer on vFairs/Airmeet is the right scope.
How long does it take to have a branded virtual event platform running?
Reselling vFairs or Airmeet with your brand: 1–2 weeks (mostly DNS and API setup). The Lovable DIY AI matchmaking layer: 1 weekend. A production RapidDev AI layer with custom sponsor analytics and branded portal: 4–6 weeks. Full custom build: 25–40 weeks.
Can RapidDev build this for my agency?
Yes — we've shipped 600+ applications and have built multi-tenant event platforms with AI matchmaking, sponsor lead extraction, and real-time transcription. Our standard AI-layer build on vFairs or Airmeet runs $20,000–$40,000 and includes all the pipeline components described on this page. Book a free 30-minute consultation at rapidevelopers.com to discuss your event calendar and client mix.
What's the real AI cost per virtual event?
For a typical 1,000-attendee, 2-day conference with 5 simultaneous tracks and 30 sponsor booths: text-embedding-3-small matching $0.01, GPT-5.4 mini agenda emails $0.45, Deepgram Nova-3 streaming $73.60 (5 tracks × 16 hrs × $0.46/hr), Sonnet 4.6 sponsor lead summaries $13.20 (30 booths × 20 chats × $0.022), highlight selection $0.50 — total ~$87.76. Captions completely dominate this number. Budget per-track-hour, not per-attendee.
What's the difference between this and a webinar platform?
Webinars are typically single-stream, 60-minute, presentation-format — one speaker, passive audience, Q&A at the end. Virtual events are multi-track (simultaneous sessions), multi-day, with expo halls, 1:1 networking rooms, sponsor booths, gamification, and attendee counts from 500 to 50,000. The AI stack overlaps (both need transcription and follow-up), but the infrastructure requirements are fundamentally different — which is why virtual events use vFairs/Airmeet while webinars can be built on top of Demio.
Do attendees need consent to be matched by AI?
For EU attendees, the EU AI Act Article 50 (effective August 2, 2026) requires disclosure that matchmaking recommendations are AI-generated. This is a disclosure requirement, not consent — attendees don't need to opt in, but they must know. For sponsor lead extraction from booth conversations, EU GDPR requires explicit consent because you're processing conversation data for commercial purposes. Add both disclosures to your registration flow and booth entry screen.
Can I build the networking matchmaking without buying vFairs or Airmeet?
Yes — the matchmaking pipeline (embeddings + cosine similarity + Supabase pgvector) is independent of the event infrastructure. You can run matchmaking on any attendee list. But the networking experience (virtual rooms, 1:1 video chat, session streaming) requires infrastructure that costs $150,000+ to build from scratch. For most agencies, embedding the AI matchmaking into an existing vFairs or Airmeet deployment is the right starting point.
Want the production version?
- Delivered in 4–6 weeks for AI matchmaking layer on vFairs/Airmeet
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.