XStudy AI Labs
Investor Relations · Confidential
Enter access code

Tomar 0.7 RAW · Live
Q3 2026 · Confidential

No.1 Code-Based
Video Generation

The next trillion-dollar shift in human communication is already happening — knowledge is moving from text to visual. But every AI video platform built to serve that shift has a fatal flaw: you cannot trust what they produce. Numbers drift. Charts are fabricated. Data visualizations are unverifiable. XStudy AI Labs built Tomar to solve this from first principles — a code-based architecture that makes hallucination structurally impossible in the rendered output. Its core market is knowledge video: education, enterprise training, and science communication. Not a better diffusion model — a different category entirely.

Zero
Hallucination in Output
by Design
Deterministic code output · no hallucinated text, charts, or data · every frame verifiable
$33B
Serviceable
Addressable Market
Verifiable video across 5 regulated verticals · 0% currently served by AI video
Up to 95%
Lower Price
vs. Veo 3.1 Standard
Tomar 0.7 RAW: priced from $0.10/10 sec · cost as low as $0.01 · vs. Veo 3.1 up to $4.00 · May 2026
$172.5M
2028E Revenue
Projection
~450% CAGR · Financial · Medical · Legal · Enterprise Training · Meeting
They're Faking Physics
We're Running It
Cost / 10 sec
$0.10+
vs $4.00 Veo 3.1 Standard
Architecture
Code
Deterministic · zero hallucination in the rendered output
SAM
$33B
Verifiable video · 0% served by AI
Pipelines
5
All verticals operational
The Investment Thesis
AI-generated video has a structural trust problem — and an entire category of high-value markets cannot unlock until it is solved.
Medical
A hospital can't sign off on clinical training video if one anatomical frame could be wrong — that's not a quality issue, it's a malpractice liability. Diffusion can't produce verifiable output; Tomar can.
Financial
An earnings report visualization where one number hallucinates is not a bad output — it is an SEC violation. Institutional publishers and financial media that need verifiable content cannot rely on diffusion-based generation.
Enterprise · Education · Research
Defense, pharma R&D, academic publishing — all need auditable, verifiable output. Diffusion produces plausible content; these markets need correct content. The distinction is categorical.
Veo 3.1 at $4.00/10 sec still cannot deploy in these markets. Tomar — from $0.10, cost as low as $0.01 — can. This is not an edge in an existing market; it is a category with no incumbent.

Today's generative video models are powerful but fundamentally undeployable for anything that requires precision — numbers, text overlays, data visualizations, scientific diagrams. This is not a model quality issue. It is a structural limitation of diffusion architectures that no amount of compute or fine-tuning can resolve.

XStudy AI Labs has built Tomar — the knowledge video compiler. Where LaTeX compiles structured text into verifiable PDFs, Tomar compiles structured knowledge into verifiable video. Input: data, formulas, simulation output, transcripts. Output: deterministic, auditable MP4 — every visual element traceable to its source. Three properties make this categorically different from every diffusion model on the market.

Zero
Hallucination · By Architecture
Deterministic code execution means every frame is verifiable before it ships. Not a filtering layer on top of diffusion — the hallucination vector does not exist in the rendered output by construction. This is the technical property that makes Tomar enterprise-deployable where no other AI video platform can be.
Up to 95%
Lower Price · Best-in-Class Margin
Priced from $0.10 per 10 seconds versus $0.50–$4.00 for Veo 3.1 — the lowest in the category. Because code-based rendering uses a fraction of the compute diffusion inference requires, Tomar produces at a cost as low as $0.01 per 10 seconds. The result: the most competitive price in the market and the strongest unit economics behind it.
Trace
Low-Cost Edits
Every frame maps to the code that produced it, so you can change one element and keep the rest. With diffusion, changing one detail means regenerating the whole clip — often altering the parts you wanted to keep. Tomar makes edits precise and cheap, with no prompt-and-pray re-rolls.
A New Market Category:
Verifiable AI Video
The existing AI video market targets creative and entertainment use cases. The market we are entering has a different requirement entirely: outputs must be auditable, defensible, and correct. This is not a feature upgrade — it is a categorical distinction that defines who can enter regulated, high-stakes verticals.
Near-Term · Platform
Unlock the
Deployability Gap
Every high-value professional market — medical, financial, legal, regulated enterprise — is blocked from adopting AI video today. Not because the tools are too expensive. Because they cannot be trusted. Tomar is the first platform that crosses the trust threshold these industries require.
LIVE PIPELINES
Financial · Enterprise Training & Education · Meeting · Legal · Medical
Mid-Term · Infrastructure
The Rendering
Engine Layer
Not selling to end users — powering the platforms that do. Conferencing platforms embed verified meeting summaries. Bloomberg renders data visualizations. Medical journals produce animated diagrams. Tomar becomes the infrastructure layer: invisible, embedded, with switching costs that compound as VideoSpec becomes the standard format for knowledge video.
PLATFORM TARGETS
Conferencing · Bloomberg · Academic Publishing · Medical
Long-Term · Paradigm
Visual as the
New Text
Human knowledge transmission is moving from text to visual. Today, producing visual content costs orders of magnitude more than writing text. When Tomar matures, that gap closes. Every piece of knowledge will naturally have a visual form — the way every document naturally has a text form today.
MARKET SHIFT
Text → Visual · Every domain · Global scale
Market Sizing
TAM · SAM · SOM
TAM
$882B
Total knowledge-video spend, led by education & training ($759B of $882B), plus medical, financial, legal, and meeting content.
SAM
$33B
The verifiable-output subset — where a single error carries educational, clinical, financial, or legal consequence.
SOM
$2–5B addressable · $172.5M 2028E target
Three pipelines live today (Education & Training, Financial, Meeting). $172.5M by 2028E ≈ 5% of the initial addressable market.
SAM Breakdown — $33B Verifiable Video Market
$12B
Education & Training
Curriculum, L&D, science comms, creators
$9B
Medical
Clinical education, surgical training, CME
$6B
Financial
Earnings, IR, regulatory filings
$4B
Legal
Compliance training, contract visualization
$2B
Meeting Intelligence
Decision documentation & review
$33B · 0% served by AI today · Tomar is the only platform that can enter
$12 + $9 + $6 + $4 + $2 = $33B
Sources: Grand View Research; HolonIQ; SkyQuest; IMARC; Josh Bersin (2025–26). TAM = addressable content spend; SAM = verifiable-output subset.
Client. Server. Render.
One Deterministic System.
Enterprise-grade infrastructure, purpose-built for deterministic and auditable video at scale.
Client Ingress ◆ Multi-Agent Orchestration Render Engine Data & State External Model Services AdminWeb Console StudioDesktop Studio API ContractShared Contract Control APIAPI Gateway Ingestion Enginedocument ingestion ◆ Agent Orchestrator plans the run · routes each step · supervises · retries on gate failure Contributordrafts content Directorplans the scene Verifierverifies facts Knowledgeretrieves evidence Evaluationscores quality Coder · Repairgenerates & repairs code ◆ Model Gateway single auditable egress · route key per agent · every model call logged Render Orchestratortimeline assemble · mux plan Hyperframes Engineframe capture Animation Engineanimation engine 3D Engine3D engine TTS Enginenarration synthesis Mux Enginecut · concat · encode Final Artifact verified MP4 · HTML Object Storageobject storage Knowledge IndexVector Index Memory + AuditLogfull agent trace Web Search APIlive evidence retrieval LLM + Embeddingcloud or on-prem HTTPS trigger run persist route key audit trace chat + embed fresh data render plan persist final artifact
Platform Products &
Revenue Streams
Core Platforms
The core technology platform powering the entire XStudy AI ecosystem — No.1 code-based AI video generation infrastructure.
AI Video · Enterprise
Tomar
Live Demos
Education Agent · STEM & Science
Financial Agent · XBRL Earnings Report
Training Agent · Corporate L&D
Enterprise Agent · Meeting Analysis
Cost Benchmark →
Five Verticals · Led by Education, Training & Science
Tomar's core market is knowledge video — education, enterprise training, and science communication. The same engine extends to meetings, finance, legal, and medical, where one wrong frame carries legal or clinical consequence. Each vertical has a dedicated pipeline, industry-standard data adapters (PPTX, SCORM, XBRL, HL7, LegalDoc), and domain-specific Scene components.
Education · Training · Science Communication
Education & Training
Tomar's core market: curriculum, enterprise L&D, science-popularization explainers, and technical training — for schools, universities, creators, and enterprise.
Precision RequirementEvery concept accurate & repeatable
Data StandardPPTX · SCORM · Simulation JSON
Pipeline Status✅ Live — 4 Industry Presets
Meeting Intelligence · Decision Fidelity
Meeting Intelligence
Meeting recordings compiled into structured 3–5 min decision-review video — speaker separation, decision extraction, action items, clip montage. 18 templates live.
Precision RequirementDecisions must be faithfully reconstructed
Data StandardMP4 · VTT · SRT · Conferencing APIs
Pipeline Status✅ Live — 18 Templates
Financial · Data Precision
Financial Reporting
XBRL-driven earnings reports, P&L waterfalls, and investor briefings — every number bound to source data. SEC- and FINRA-aligned. 11 Scenes live.
Precision RequirementNumbers must be exact
Data StandardXBRL · SEC EDGAR · US GAAP
Pipeline Status✅ Live — 11 Scenes
Medical · Clinical Accuracy
Medical & Clinical
Drug-mechanism animation, clinical imaging, anatomical walkthroughs, and patient education — every conclusion traceable to source, HIPAA-aware.
Precision RequirementConclusions must be clinically accurate
Data StandardHL7 FHIR · PDB/CIF · Clinical Guidelines
Pipeline Status✅ Live
Legal & Compliance · Textual Completeness
Legal & Compliance
Contract-clause visualization, regulatory comms, red-line comparison, and compliance training — with word-for-word completeness verification.
Precision RequirementText must be word-for-word complete
Data StandardLegalDoc · ComplianceAdapter
Pipeline Status✅ Live
The Market No Lab
Will Prioritize
Cost is not what separates Tomar from incumbents. Architecture is. Diffusion models are probabilistic by design — they cannot produce outputs that are auditable or verifiable, regardless of how much compute you apply.
Tomar
Veo 3.1
Lite · Fast · Standard
Runway Gen-4.5
Kling 3.0
Pika 2.5
Architecture
Code-Based
Diffusion
Diffusion
Diffusion
Diffusion
Output Verifiable
Physics-Accurate
Traceable Edits
Cinematic Quality
Not focus
~
Cost / 10 Sec
from $0.10
$0.50–4.00
$0.78–2.00
$0.53–0.70
$1.42
Max Duration
10 min
60s
40s
15s native
3 min via extend*
15s
*Kling reaches longer durations only by iterative extension (续写) — stitching successive short clips — not a single coherent generation. Tomar's 10 min is one continuous, deterministic render.
The Honest Risk — and Our Real Moat
Why won't the labs
build this first?
Google · Veo 3.1
Diffusion Path Dependency
Billions sunk into Veo's diffusion stack — Google won't write it off while revenue grows. Its priority is YouTube and Workspace engagement, not verifiable output for a hospital.
Anthropic · Claude
Bandwidth and Will
Anthropic is all-in on LLMs, safety, and Claude's coding products. Verifiable enterprise video isn't on their roadmap — the opportunity cost is too high while the LLM market is still being won.
OpenAI / Runway / Kling
Wrong Optimization Target
They optimize for "does this look cinematic?" Tomar optimizes for "is this frame legally defensible?" Different talent, data, and go-to-market — and a diffusion model can't be retrained into code. It has to be rebuilt.
Technical Wedge · Today
Hybrid Rendering Architecture
Code for verifiable precision, generative models for aesthetics, AI for scale. A paradigm, not a feature — and years to operationalize across regulated domains.
Durable Moat · Data Flywheel
Vertical Domain Depth
Every deployment deepens domain knowledge — clinical protocols, SEC chart formats, regulatory structures. Not generic training data; institutional knowledge that compounds with each customer.
Structural Moat · Switching Cost
Workflow Integration Lock-In
Once embedded in a hospital's documentation or a bank's reporting workflow, the switching cost isn't financial — it's regulatory. The integration is the product.
Why This Cannot Be Fixed
Diffusion models are probabilistic by nature — they generate visually plausible outputs through iterative noise sampling. This is not a limitation that more compute or training data resolves. The verification problem is architectural: there is no deterministic path from input to output, and therefore no auditable record of how any frame was produced. Tomar's code-based approach inverts this entirely. Every frame is the output of executable logic — inspectable, reproducible, and defensible. This distinction is not a competitive edge. It is a categorical boundary between markets Tomar can serve and markets no diffusion-based platform ever will.
Cost Advantage and
The Deployability Advantage
Press Release
Tomar is launching a new type of video generation. We will transfer information with a revolutionary way that will change the method of learning physics and mastering knowledge, pushing forward a much better world.

Tomar 0.7 RAW is priced from $0.10 per 10 seconds — the lowest in the category — and produced at a cost as low as $0.01, the strongest unit economics among AI video platforms. But the more important number is not the price gap. It is the deployability gap: diffusion models at any price cannot be used in a hospital, an investment bank's reporting workflow, or a defense contractor's documentation pipeline. Tomar can.

Zero Hallucination in the Rendered Output. Tomar's code-based architecture produces deterministic, auditable frames — every element placed by executable logic, not probability. Content passes verification gates before render, and the render itself cannot introduce hallucination — a property no diffusion model can replicate regardless of training scale.

Traceable, Low-Cost Edits. Because every frame maps to the code that produced it, a user can trace any change to the exact element and edit just that part — keeping everything else untouched. With diffusion, changing one detail means regenerating the whole clip, often altering the parts you wanted to keep and forcing repeated re-rolls. Tomar makes editing precise, cheap, and predictable — far more usable than prompt-and-pray diffusion, and an audit requirement for regulated contexts.

Compute-Efficient. Code-based rendering requires a fraction of the compute that diffusion inference consumes per output — dramatically lowering the cost per second of finished video. This is what enables Tomar's production cost as low as $0.01 and its category-leading price, and lets the platform scale to enterprise volume without the inference-cost ceiling that constrains diffusion competitors.

Up to 95%
Lower Price vs. Veo 3.1 Standard
from $0.10
Price / 10 Sec · cost as low as $0.01 · Tomar 0.7 RAW
Tomar 0.7 RAW Price & Cost Benchmark
Price per 10 Sec (USD) · AI Video Generation Platforms · May 2026 · Tomar bar shows cost vs. price
Tomar 0.7 RAW
(XStudy AI)
COST · as low as $0.01
PRICE · from $0.10
Kling 3.0
(Standard)
Kling 3.0
(Pro)
Runway Gen-4.5
(Standard)
Veo 3.1 Lite
(Google · 1080p)
Luma Ray3
(Standard)
Pika 2.5
(Standard)
Veo 3.1 Fast
(Google · 1080p)
Veo 3.1 Standard
(Google · 1080p)
Sora 2 Pro HD
(Shut Down Mar 2026)
XStudy AI (Tomar)
Competitors
Source: XStudy AI Labs internal cost benchmark, May 2026. Competitor pricing based on published API rates and official documentation. Tomar: proprietary code-based synthesis architecture.
Executive Team
Co-Founder & CEO
Bill Wu
Bill Wu co-founded XStudy AI Labs and leads the company's strategic direction, enterprise partnerships, and go-to-market efforts. He was named to Forbes' 100 Most Influential Chinese list and recognized as an Asia-Pacific Leader Under 30. Previously, he served as Editor-in-Chief of the Silicon Valley Business Review and as a Product Manager at Waterdrop Inc. (NYSE: WDH). Bill is driving Tomar's deployment across regulated enterprise verticals and managing key enterprise and platform partnerships.
Forbes 100 Most Influential Chinese APAC Leaders Under 30 NYSE: WDH Alumni UNC Chapel Hill
President & COO
Jack Feng
Jack Feng serves as President & COO of XStudy AI Labs. As a founding engineer and senior architect at WebEx, he led the development of its teleconferencing and data communication platforms. He later founded Cenwave and served as its Chairman and CEO until its acquisition by Huawei, after which he became Vice President of Huawei's Unified Communications and Collaboration product line, overseeing product management and marketing engineering for video and collaboration products serving global enterprises. Over three decades, Jack has built and scaled video infrastructure, and that expertise now directly powers Tomar's rendering architecture, enterprise GTM strategy, and cross-vertical platform strategy. He holds a B.S. from Tsinghua University.
WebEx Founding Engineer Cenwave Chairman & CEO (acq. Huawei) Huawei VP · UC&C B.S. Tsinghua University
Co-Founder & CTO
Sean Bai
Sean Bai co-founded XStudy AI Labs and serves as Chief Technology Officer, driving the company's AI research and development across the full technical stack. Prior to XStudy AI Labs, he served as Software Director at Turing AI ($500M valuation, Hillhouse Capital-backed) and as a Software Engineer at Oracle. Sean holds an M.S. from Northwestern University. He leads the architecture of Tomar's VideoSpec engine, multi-pipeline rendering infrastructure, and enterprise AI integrations.
Turing AI (Hillhouse-backed) Oracle Alumni M.S. Northwestern
CMO
David D. Williams
David Williams has over 20 years of experience leading brand management, omnichannel marketing, and marketing operations across global CPG and B2B organizations. As Senior Director of Omnichannel Marketing at GOJO Industries — maker of PURELL® — he drove brand awareness and share growth across owned, paid, and ecommerce channels, while leading account-based marketing programs delivering 8–10x ROAS. He previously managed an $800M+ brand portfolio and a $245M P&L at S.C. Johnson. David holds an MBA from Cornell University and is a CFA charterholder.
GOJO / PURELL® Alumni S.C. Johnson Alumni MBA, Cornell University CFA
Revenue Projections
& Growth Trajectory
Revenue by Business Unit (USD Millions)
Business Unit FY 2026E FY 2027E FY 2028E Revenue Model Notes
Tomar Platform
API · Subscription
$1.4M $11.0M $43.1M API · Subscription Core generation infrastructure. Priced from $0.10/10 sec (cost as low as $0.01). Scales from SMB $10–50K/yr to institutional platform license $1–5M+/yr.
Business Solutions
Training · Education · Financial · Meeting · Medical · Legal
$4.26M $32.85M $129.4M Pilot · Annual Sub · Platform License Education $50K–10M+/yr · Enterprise training $100K–10M+/yr · Financial institutional license $1–5M+/yr · Meeting Intelligence via conferencing platforms · Medical $500K–5M/yr · Legal $500K–10M/yr.
Total Revenue $5.66M $43.85M $172.5M Multi-vertical platform ~30× growth · ~450% CAGR · Tomar 25% · Business Solutions 75%
YoY Summary
Fiscal YearTotal RevenueYoY Growth
FY 2026E$5.66M
FY 2027E$43.85M+675%
FY 2028E$172.5M+293%
3-Year CAGR$172.5M by 2028~450% CAGR
Note: Projections are forward-looking estimates subject to material risks and uncertainties. Tomar platform revenue reflects subscriber and API usage growth across five verticals. Vertical Applications revenue assumes initial enterprise licensing agreements signed H2 2026. All five vertical pipelines are operational. Actual results may differ materially.
Pilot Partners & Traction
PartnerSegmentStatus
ZoomMeeting IntelligenceIN DISCUSSION
Akrostar · Figgs GroupEnterprisePilot
NetClass Technology (NASDAQ: NTCL)Education / EdTechPilot
Western Reserve · Springfield CommonwealthEducationPilot
Live today: all five vertical pipelines operational · 18 Meeting Scene templates · 11 Financial Scenes · 4 industry training presets.
Stacked Revenue by Business Unit · 2026–2028E
$0 $43M $86M $129M $172M $5.7M 2026E $43.9M 2027E $172.5M 2028E ~30× growth
Tomar Platform
Business Solutions
CAGR ~450% · 2026–2028E · Tomar 25% · Business Solutions 75%

Disclaimer. The information presented on this page is for informational purposes only and does not constitute an offer to sell, a solicitation of an offer to buy, or a recommendation regarding any securities. Certain statements contained herein may constitute forward-looking statements based on current expectations and projections about future events. These statements are subject to risks, uncertainties, and assumptions, and actual results may differ materially from those expressed or implied. Financial projections reflect management estimates and have not been independently audited. XStudy AI Labs assumes no obligation to update any information presented herein. Prospective investors should conduct their own due diligence and consult with qualified financial, legal, and tax advisors before making any investment decision.