Pre-Seed · $250K SAFE · No Cap

“The Google Translate for how the world actually speaks.”

Every voice,
understood.

VALSEA is speech infrastructure for the accents global AI gets wrong — Southeast Asia, India, China, MENA. We route every voice to the right local engine and return structured JSON directly into your ERP, CRM, or compliance archive.

$6.9B

TAM · 2025 verified

$1.7B

SAM · SEA + India + China + Africa

$1M ARR

SOM target · Dec 2027

Request Investment Memo Schedule Intro Call →

The problem

Global speech AI fails
where most of the world lives.

Azure, OpenAI Whisper, and Google Speech were trained on North American and European accents. They hit 95% accuracy in standard English — and 60–70% on Vietnamese, Singlish, Tamil-accented English, and Mandarin-accented English. Below 80%, enterprise voice automation is worthless.

~65%

Accuracy on Vietnamese

Azure and OpenAI Whisper on standard Vietnamese speech — below the threshold for any enterprise automation workflow. Every error compounds downstream in ERP and CRM.

4.5B

People in underserved markets

SEA, India, China, MENA, Central Asia, Sub-Saharan Africa. The majority of the world — invisible to global speech AI platforms optimised for Western accents and vocabulary.

Automation value at 65% accuracy

One-in-three words wrong means manual review of every output. Voice-to-ERP requires 90%+ accuracy. The gap between global engines and that threshold is the entire market VALSEA addresses.

The approach

Orchestration, not invention.

We don’t build every model from scratch. We route each voice to the best-in-class local engine for that language, then layer context awareness and accent correction on top. The result: 10–30% accuracy lift over Azure and OpenAI, with a fraction of the R&D cost.

Voice arrives

Any input channel

WhatsApp, Zalo, IVR, call recording, live stream, dictation

→

VALSEA routes

Language → best local model → context + accent correction

VinAI for Vietnamese · iFLYTEK for Mandarin · custom for Singlish · fine-tunes for Bahasa

→

Structured output

JSON → system of record

ERP · CRM · EMR · WMS · compliance archive · QA dashboard

Language routing

Best engine per language

Vietnamese → VinAI, Bahasa → local Hugging Face fine-tune, Mandarin → iFLYTEK, Singlish → custom trained. Each voice goes to the model it deserves, not a universal model trained on American English.

Context awareness

Domain-specific accuracy

Fintech terms (CASA, FX rate), healthcare drug names, logistics SKU codes, legal terminology — context layers prevent generic transcription errors that corrupt downstream data and trigger compliance failures.

Accent correction

Dialect-aware post-processing

“Chicken lies rice” → “chicken rice.” Singlish particles, Vietnamese tones, Indian-English retroflex consonants. The last 5% accuracy gap that kills enterprise automation, closed by accent-aware correction.

Competitive advantage

Why VALSEA wins in the markets
Azure and Google ignore.

Global players optimise for Western markets and leave 65% accuracy gaps in Southeast Asia, India, and China. VALSEA is purpose-built for the accuracy threshold that makes enterprise automation possible in those markets.

01 · Accuracy

10–30% better than Azure & OpenAI on Vietnamese

Partnership with local providers training speech models on NVIDIA GPUs under NVIDIA’s on-premises exception program for Vietnam’s data sovereignty regulations. Access to training data and compute infrastructure unavailable to global cloud providers subject to export controls.

NVIDIA on-prem exception program

02 · Routing

Best-in-class model per language, not one universal model

Global engines use a single model for all languages. VALSEA routes each language to the highest-performing local model available. The routing engine — language detection, model selection, confidence scoring, fallback logic — is the core IP and the competitive moat.

Language-aware routing engine

03 · Integration

API-first JSON output, directly into ERP / CRM / compliance

Not just transcription — structured data extraction. A voice note saying “100 boxes SKU-4421 deliver to Jurong by Thursday” becomes a structured order record in SAP. The gap between transcription and enterprise automation is the integration layer VALSEA owns.

JSON → ERP / CRM / EMR / WMS

04 · Infrastructure vision

Speech OS for Asia — Elston’s ML research roadmap

Elston’s ML research roadmap targets full global accent coverage — every language, every dialect. The long-term aspiration: foundational speech infrastructure any platform can build on, the same way Google Translate became infrastructure for every language app. Series B upside not in SOM.

Elston ML research · global accent roadmap

Market opportunity

$6.9B market. Every enterprise
that touches non-English voice.

The TAM is not a niche — it’s the full global enterprise speech infrastructure market. US companies running contact centers in Manila. European MNCs with back-office teams in Bangalore. Chinese manufacturers managing distributors in Ho Chi Minh City. Any business that processes voice in a language Azure, Google, or OpenAI gets wrong is a VALSEA customer. The SAM is the non-English, non-North-American slice of that market — the fastest-growing portion, and the one every incumbent systematically underserves.

TAM · 2025

$6.9B

Global enterprise speech infrastructure. Any business processing non-English voice is in this market — US and EU MNCs included. Enterprise STT API $4.66B (Fortune BI) + Contact Center Voice QA $2.25B (Credence Research). Both verified 2025.

SAM · geographic rollout

$1.7B

The non-English, non-North-American slice of the $6.9B market. SEA + India + Gulf + Africa + MENA — reachable with current language coverage and go-to-market. Derived from verified APAC/MENA regional splits (Credence Research) + non-English ASR proxy (Fortune BI).

SOM · Dec 2027

$1M ARR

Bottom-up: 22 B2B accounts ($700K) + 1,450 Memra subscribers ($301K). 0.014% of SAM — SEA only, cross-border corridors not counted.

Why VALSEA wins · the corridors incumbents cannot serve

The $6.9B market has a structural fault line: North American incumbents (Azure, Google, OpenAI) are optimised for English. The fastest-growing demand is in the opposite direction — US and EU MNCs expanding into SEA and India, Chinese manufacturers managing global supply chains, Indian diaspora running commerce across the Gulf. These are not niche corridors. They are the new centre of gravity in global trade, and every one runs on voice that the incumbents cannot accurately transcribe.

Corridor 1 · China → SEA

$700B+ annual trade

Guangdong and Fujian manufacturers sell to Vietnamese factories, Indonesian distributors, and Thai retailers. Mandarin ↔ Vietnamese ↔ Bahasa ↔ Thai. Zero English in the supply chain. Chinese factory reps manage SEA sales teams entirely in local languages.

Mandarin ↔ Vietnamese Mandarin ↔ Bahasa Cantonese ↔ Thai

Corridor 2 · China → Africa

$282B trade · 2,500+ firms

2,500+ Chinese companies operate in Sub-Saharan Africa. Construction, telecoms (Huawei, ZTE), mining, and infrastructure. Chinese site managers in Nigeria, Kenya, Ethiopia, DRC communicate daily with local supervisors. Mandarin ↔ Swahili ↔ Yoruba ↔ Amharic. No viable AI bridge exists.

Mandarin ↔ Swahili Mandarin ↔ Yoruba Mandarin ↔ Amharic

Corridor 3 · India ↔ Gulf

8M+ diaspora · $100B remittances

8M+ Indians in UAE and Saudi Arabia. Gujarati textile and diamond merchants in Dubai run wholesale empires: Arabic wholesale buyers call in Arabic, Gujarat suppliers call back in Gujarati, logistics partners speak Tamil. Three languages, zero English, billions in commerce. Banks, insurers, and logistics firms serving this corridor have no AI speech layer.

Arabic ↔ Gujarati Arabic ↔ Tamil Arabic ↔ Malayalam

CPEC · China ↔ Pakistan

$62B corridor

China-Pakistan Economic Corridor. Chinese engineers in Karachi, Lahore, and Gwadar managing infrastructure with local crews. Mandarin ↔ Urdu ↔ Punjabi ↔ Pashto across every project site. BRI’s flagship corridor, zero speech AI coverage.

Mandarin ↔ Urdu Mandarin ↔ Punjabi

Philippines ↔ Gulf (OFW)

2.2M workers · $10B remit.

2.2 million Filipinos in UAE, Saudi Arabia, and Qatar. Workers negotiate contracts, resolve HR disputes, and file legal claims entirely over voice calls. Tagalog ↔ Arabic. Employers, banks, and legal services have no compliant speech layer for OFW interactions.

Tagalog ↔ Arabic Tagalog ↔ Gulf English

Bangladesh ↔ China (RMG)

$6B+ fabric supply chain

Bangladesh is the world’s second-largest garment exporter. Chinese fabric mills in Guangdong and Zhejiang supply Dhaka factories entirely over voice — order negotiation, spec changes, shipping confirmations in Bengali ↔ Mandarin with no interpreter layer. Errors mean rejected shipments worth millions.

Bengali ↔ Mandarin

BRI Central Asia ↔ China

$50B+ · 5 countries

Kazakhstan, Uzbekistan, Kyrgyzstan, Tajikistan, Turkmenistan along the New Silk Road. Chinese infrastructure and energy companies operate jointly with Central Asian partners. Mandarin ↔ Kazakh ↔ Uzbek. Fastest-growing BRI segment, entirely underserved by speech AI.

Mandarin ↔ Kazakh Mandarin ↔ Uzbek

Mekong · Myanmar/Laos ↔ China

$30B+ Mekong trade

Kunming-Singapore Railway, Laos-China Railway, and Mekong waterway trade. Chinese merchants trade with Myanmar jade dealers, Lao rubber farmers, and Cambodian garment factories entirely in local languages. Burmese ↔ Mandarin, Lao ↔ Mandarin — no AI solution exists for either pair.

Burmese ↔ Mandarin Lao ↔ Mandarin

India ↔ Africa

$100B trade · 3M diaspora

3 million Indians in East and Southern Africa — Kenya, Tanzania, Uganda, South Africa — running businesses that connect Gujarat, Mumbai, and Chennai suppliers to African distributors. Hindi/Gujarati ↔ Swahili ↔ Zulu. Indian pharmaceutical and FMCG companies distributing across Sub-Saharan Africa face the same zero-coverage problem in every sales call.

Hindi ↔ Swahili Gujarati ↔ Zulu

TAM breakdown · global enterprise speech infrastructure · 2025 actuals

Sub-market	2025 size	CAGR	Source (verified URL)	VALSEA fit
Enterprise STT / ASR API Speech-to-text API delivered to enterprise; ASR, speaker ID, language detection	$4.66B	20.7% CAGR	Fortune Business Insights — Speech-to-Text API Market, 2025 fortunebusinessinsights.com/speech-to-text-api-market-102781	Core revenue line — every VALSEA API call is an ASR transaction
Contact Center Voice QA & Compliance Call recording analysis, QA scoring, regulatory archiving — BPO and enterprise CC	$2.25B	8.4% CAGR	Credence Research — Contact Center QA Software Market, 2024 credenceresearch.com/report/contact-center-quality-assurance-software-market	MAS compliance, BPO QA, call archiving — APAC 21% share of global market
Total TAM	$6.9B	—	Both numbers confirmed at live URLs, June 2025. This is a global market — US and EU enterprises running non-English voice operations are already in it. VALSEA’s wedge is the non-English portion that incumbents underperform on: APAC, MENA, India, and cross-border trade corridors.

SAM · geographic rollout sequence

Region	Entry	SAM	GTM strategy
Southeast Asia VN, ID, SG, MY, PH — languages live or in progress	Q2 2026	$0.47B	Direct outbound + Vietnamese ASR Distributor channel. APAC voice QA: 21% × $2.25B Credence Research.
China MNC-only via Hangzhou entity · ICP + data localisation	Q3 2026	$0.17B	MNC BPO and CX teams only · Alibaba Cloud marketplace. China share ~3.7% of non-English ASR proxy.
India Hindi ASR beta · distribution partnership in progress	Q4 2026	$0.38B	AWS India marketplace + distribution partner. Non-English ASR proxy: $4.66B × (1−32.3%) × ~12%.
Gulf · Indian diaspora corridor 8M+ Indians in UAE + Saudi · Gujarati / Tamil / Malayalam merchants ↔ Arabic buyers	Q4 2026	$0.28B	Cross-border corridor proxy: ~15,000 corridor enterprises × $19K avg ARPU. India remittance $129B (World Bank 2024) evidences corridor scale.
Sub-Saharan Africa · Chinese ops 2,500+ Chinese firms in KE/NG/ET/ZA/AO/DRC · Mandarin ↔ Swahili/Yoruba/Amharic	Q1 2027	$0.27B	China-Africa $275B trade (BU Global Development Policy Center 2024). 2,500+ firms × $107K avg speech infra spend proxy.
MENA (Arabic dialect routing) Arabic + Gulf enterprise CX · research roadmap Q2 2027	Q3–Q4 2027	$0.13B	MENA voice QA: 6% × $2.25B (Credence Research MENA split). Arabic dialect routing unlocks Gulf enterprise CX + BPO at scale.
Total SAM		$1.7B	Excludes Japan, Korea, ANZ — established incumbents. Derived from verified 2025 market data: Fortune BI STT API + Credence Research APAC/MENA regional splits. Cross-border corridor component uses proxy ARPU model, not extrapolated market reports.

Market data sources · all numbers verified at live URLs

[1] Fortune Business Insights — Speech-to-Text API Market: $4.66B (2025), 20.66% CAGR, NA 32.27% share
fortunebusinessinsights.com/speech-to-text-api-market-102781

[2] Credence Research — Contact Center QA Software: $2.25B (2024)→$4.29B (2032), APAC 21%, MENA 6%
credenceresearch.com/report/contact-center-quality-assurance-software-market

[3] World Bank Blog — Remittances 2024: India $129B, Philippines $40B, Global $905B
blogs.worldbank.org/en/peoplemove/in-2024--remittance-flows-to-low--and-middle-income-countries-ar

[4] BU Global Development Policy Center — China-Africa trade $275B (2024)
bu.edu/gdp/2026/05/20/what-new-data-reveals-about-china-africa-economic-ties

[5] PRNewswire / China Report — China-ASEAN trade $982.34B (2024)
prnewswire.com/apac/news-releases/top-10-news-stories-on-2024-china-asean-cooperation-302394433.html

[6] MarketsandMarkets — Global speech & voice recognition: $8.49B (2024), $9.66B (2025), 19.1% CAGR
marketsandmarkets.com/Market-Reports/speech-voice-recognition-market-202401714.html

All sources navigated and verified June 2025. Numbers are current-year actuals or near-term forecasts, not decade-out projections. Cross-border corridor trade values evidence market opportunity but are not claimed as TAM — no verified speech-AI revenue data exists for this specific category.

SOM · bottom-up to $1M ARR by Dec 2027

B2B segment	Accts	ACV	ARR
BPO / CX SG + VN + ID, incl. 1 China MNC	3	$80K	$240K
ISV Partners AWS · GCP · Alibaba · MSFT via distributor	5	$40K	$200K
CX Integrators HubSpot · Salesforce · Zendesk	2	$40K	$80K
Memra B2B Conference orgs, tuition centres, ops teams	12	$15K	$180K
Total B2B	22	$32K avg	$700K

B2C (Memra app)	Subs	Price	ARR
Pro tier Business travellers, professionals	600	$29/mo	$209K
Personal tier Tourists, students, general audience	850	$9/mo	$92K
Total B2C	1,450	—	$301K

Grand total · Dec 2027 (Month 18)

$1,001K ARR

22 B2B accounts + 1,450 B2C subscribers = 0.014% of SAM. SEA-only — China, India, Africa, and cross-border corridor revenue not counted in SOM.

Series B vision · The Grab model

Platform tax multiplier:
Speech OS for South-South trade

The world’s fastest-growing trade corridors — China-SEA ($700B+), China-Africa ($282B), India-Gulf ($100B+) — all run between non-English-speaking markets. These corridors need a speech infrastructure layer the way global trade needed SWIFT. VALSEA’s hackathon program seeds products built on this API: cross-border order automation, multilingual compliance archiving, diaspora banking CX. Elston’s ML research roadmap targets full language coverage for every corridor. Grab built ride-hailing, then became Southeast Asia’s OS. VALSEA builds speech routing, then becomes the voice infrastructure for the non-English world. Series B upside not in SOM.

Ecosystem play

Grab model

Global accents

Elston ML research

Series B upside

Not in SOM

Traction

Enterprise proof points
and pipeline.

POC → Paid conversion

Coca-Cola APAC

Flagship enterprise logo · SG + VN operations

Vietnamese ASR for agent call transcription and QA automation across SG and VN operations. POC completed. Commercial contract in final negotiation. $70–90K ACV target. Proof point for enterprise sales motion and accuracy advantage narrative.

Pipeline · API integration scoped

Kyanon Digital Vietnam

Loyalty app platform · Fortune 500 end clients

Kyanon builds loyalty and CX apps for Fortune 500 companies across Vietnam. VALSEA speech API integration scoped post-Coca-Cola — voice-driven loyalty interactions and agent workflows. Channel entry to Fortune 500 client base without direct enterprise sales cycles.

Channel partner · Pre-qualified pipeline

Vietnamese ASR Distributor

In-country distribution · NVIDIA program access

Local distribution partner with GPU access under NVIDIA’s on-premises exception program for Vietnam. Provides pre-qualified enterprise leads, in-country sales support, and model training infrastructure. Competitive moat against global players subject to export controls.

Pipeline · AWS ISV program

AWS Marketplace ISV listing

Cloud marketplace channel · procurement simplification

VALSEA on AWS Marketplace removes procurement friction for enterprise buyers with existing AWS commitments. ISV program provides co-sell support and access to the AWS partner network across SEA. Targeting $200K ARR via cloud channel. GCP and Alibaba Cloud listings in parallel.

The ask

$250,000

20% SAFE · No Valuation Cap · Converts at Series A pricing.
First close targeting July 2026 — 2 spots remaining.

$250K

Raise amount

20%

SAFE equity

No cap

Converts at Series A

Use of funds

Engineering — ML infra, API, deployment

40%

GTM — enterprise AE, conferences, channel

30%

Compute — GCP training, inference hosting

20%

Legal & IP — patents, contracts, SOC 2

10%

SOM · 18-month path to $1M ARR

Tier	Accts	MRR	ARR
Enterprise (~$8K/mo)	4	$32K	$384K
Scale ($2,500/mo)	3	$7.5K	$90K
Growth ($499/mo)	12	$6K	$72K
Starter+ ($299/mo)	15	$4.5K	$54K
Starter ($149/mo)	20	$3K	$36K
Developer ($49/mo)	22	$1K	$13K
API SUBTOTAL		$54K	$648K
Memra Pro Teams	20	$2.6K	$31K
Memra Enterprise	5	$2.3K	$27K
Memra Business ($52)	28	$1.5K	$17K
Memra Pro ($29)	600	$17.4K	$209K
Memra Personal ($9)	400	$3.6K	$43K
MEMRA SUBTOTAL		$27K	$328K
TOTAL (Month 18)		$81K	~$1M

Jul ’26 → Dec ’27 · 65/35 API/Memra · 4 enterprise deals = 46% of ARR · SEA only, corridors not counted.

Request Investment Memo Schedule Intro Call →

Enterprise use cases

27 workflows VALSEA
solves out of the box.

Across 13 verticals, 9 trade corridors, and 20+ language pairs — voice data flows through all of these workflows today, lost in manual transcription and re-keying. VALSEA routes every voice note, call, or dictation to the right local engine and delivers structured output directly into the system of record.

Supply chain & logistics

Supply chain

FMCG distributor order capture

Suppliers send SKU + quantity + delivery address via WhatsApp or Zalo voice notes in Vietnamese or Bahasa. Agents manually listen and re-key into SAP — 10,000+ daily orders across SEA FMCG chains. One mishearing corrupts the whole order.

WhatsApp / Zalo voice→Structured JSON → SAP

Vietnamese tone correction + SKU context model. Eliminates manual re-keying entirely.

Logistics

Singlish last-mile logistics

Warehouse ops at Pandan Loop and Tuas send delivery instructions via voice note. Global engines mishear Singlish address particles and Singapore street names consistently, creating failed deliveries and driver disputes.

Telegram voice note→Delivery record + WMS

Singlish dialect model — SG address, street code, and postal accuracy.

Financial services & insurance

Financial services

MAS compliance call transcription

MAS-regulated advisory firms must archive every client call verbatim. Current tools fail on Singlish code-switching and financial terminology. Compliance teams manually review failures — missing the 100% coverage mandate.

Call recording→Compliance archive + flag

Fintech terminology + Singlish. Regulatory pull — not discretionary ROI.

Insurance

Field claims intake

Field agents record voice notes at accident scenes in Tamil, Bahasa, or accented English. Admin manually fills structured claims forms — 45+ minutes per claim. High volume across SG, MY, and ID insurers. Errors trigger re-inspection cycles.

Agent voice (on-site)→Pre-filled claims form

Insurance entity extraction + multilingual routing. 45 min → 5 min per claim.

Healthcare

Doctor dictation → EMR

Doctors dictate clinical notes in accented English or local language post-rounds. Dragon Medical fails on SEA-accented speech and local drug trade names. VALSEA routes to the right local model with medical entity extraction, reducing dictation time and eliminating transcription errors that affect patient records.

Doctor dictation→Structured EMR entry

Medical terminology model + accent correction for SG/MY/VN hospital groups.

Healthcare

Patient intake & triage hotline

Hospital hotlines handle patients calling in Bahasa, Tamil, and Mandarin mixed with English. Agents mishear symptoms and triage escalation triggers. Misclassified triage calls lead to delayed care and potential liability for the hospital operator.

Inbound patient call→Triage score + ticket

Real-time transcription + symptom entity extraction. Reduces escalation errors.

HR & talent

Cross-border recruitment screening

Pan-SEA employers screen candidates from Vietnam, Indonesia, and India over phone. Interviewers struggle with heavy accents — introducing bias — or make costly interpreter bookings for every screening call, adding cost and friction to high-volume hiring.

Screening call→Transcript + scorecard

Multi-accent routing (VN/ID/IN) + competency keyword extraction and scoring.

HR & talent

Factory floor voice attendance reporting

Vietnam and Indonesia factory supervisors submit daily attendance and overtime verbally to a phone line or WhatsApp. Data is manually typed into payroll and MES systems — delayed 12–24 hours and error-prone on employee names and IDs.

IVR / WhatsApp voice→Payroll / MES entry

Vietnamese/Bahasa + HR entity extraction (employee ID, hours, shift code).

Field sales & FMCG

Field sales

FMCG field rep order taking

Unilever, Nestlé, and P&G reps visit 20+ retail outlets daily, taking restocking orders verbally in local language. Orders noted on paper, re-keyed into CRM after hours — losing product detail and momentum, and inflating order error rates.

Rep voice note (in-store)→Live CRM order entry

Product name + SKU context model in local language. Removes end-of-day re-keying entirely.

Real estate

Property agent CRM logging

SG and MY real estate agents record voice memos after viewings — property specs, client preferences, price feedback. Logged manually hours later, losing granular detail and slowing follow-up speed in a competitive same-day market.

Post-viewing voice memo→CRM record + follow-up

Property entity extraction (unit, price, developer, client) + Singlish accuracy.

Cross-border trade corridors

China → Africa

Chinese construction firm ↔ African site supervisors

CCECC, Sinohydro, and CRBC project managers in Nigeria, Kenya, and Ethiopia communicate daily with local supervisors in Swahili, Yoruba, and Amharic. Safety instructions, work orders, and materials requests flow via voice notes. Mistranslation is a safety and compliance risk on regulated infrastructure projects.

WhatsApp voice (site)→Mandarin ↔ Swahili/Yoruba

First ASR routing Mandarin ↔ East/West African languages. Zero incumbent solutions exist.

India ↔ Gulf

Gujarati merchant ↔ Arabic wholesale buyer

Dubai and Sharjah textile and diamond traders from Gujarat run businesses where Arabic wholesale buyers call in Gulf Arabic, Gujarat suppliers call back in Gujarati, and logistics partners speak Tamil. Entire commercial operations run across three languages with zero English. Banks and trade finance providers serving this corridor have no speech data layer.

Phone call→Arabic ↔ Gujarati / Tamil

Gulf Arabic + Gujarati dialect routing. Unlocks trade finance, CX, and compliance for $50B+ corridor.

China → SEA

Shenzhen OEM ↔ Jakarta / Hanoi distributor

Electronics and FMCG manufacturers in Shenzhen and Dongguan manage distributor relationships across SEA with Chinese-speaking sales reps and Bahasa/Vietnamese-speaking buyers. Order negotiation, product spec calls, and delivery disputes happen in the supplier’s language and the buyer’s language — never English. Failed conversations cost orders.

WeChat / Zalo voice→Mandarin ↔ Bahasa / Vietnamese

Mandarin + SEA language routing. China’s $700B+ SEA trade corridor, voice-automated for the first time.

Legal & arbitration

Legal

Court & arbitration transcription

SIAC and Indonesian courts require verbatim transcripts. Parties speak accented English, Bahasa, or code-switch mid-testimony. Existing tools produce error rates requiring full manual review by legal stenographers — adding cost and delay to every proceeding.

Hearing audio→Verbatim legal transcript

Legal terminology + multi-speaker diarisation + code-switch handling across SEA languages.

Commodity trading

Commodity trade confirmation

Palm oil, rubber, and coal brokers in MY, ID, and VN confirm trades over phone in local accents. Verbal price, quantity, and grade confirmations are manually logged — disputes arise from transcription errors on amounts and delivery terms.

Broker phone call→Structured trade ticket

Financial + commodity entity extraction (grade, incoterm, price, volume, counterparty).

Government & public services

Government · Indonesia

Village administration voice filing

East Java and West Java village officers receive citizen requests in Javanese and Sundanese — not Bahasa Indonesia. Officials must manually translate and enter data into BPK-mandated national admin systems. VALSEA routes regional language speech directly into structured government forms, removing the translation layer entirely.

Citizen voice request→Javanese / Sundanese → gov form

Regional language routing for 140M+ Javanese and Sundanese speakers. First speech layer for Indonesian local gov.

Government · Philippines

BIR tax helpdesk multi-dialect logging

Philippine Bureau of Internal Revenue call centers handle taxpayer queries in Tagalog, Bisaya (Cebuano), Ilocano, and Hiligaynon. Agents manually log every call. Misclassification delays resolution by 3–5 days. VALSEA transcribes and classifies each call at intake, routing to the right resolution team automatically.

Inbound taxpayer call→Classified ticket + CRM

Philippine multi-dialect routing (Tagalog, Bisaya, Ilocano, Hiligaynon) + intent classification.

Retail & live commerce

Retail · SEA

TikTok Shop live seller inventory capture

Thai and Indonesian live commerce sellers narrate product SKUs, pricing, and variant availability during 2-hour live sessions on TikTok Shop and Shopee Live. Buyer orders and inventory updates all happen verbally in real time. No structured record until the stream ends — high fulfilment error rates without real-time ASR capture.

Live stream voice→Real-time inventory + order log

Thai and Bahasa real-time ASR + product entity extraction (SKU, qty, price, variant) during live session.

E-commerce · SEA

Marketplace seller dispute voice logging

Shopee, Lazada, and Tokopedia sellers record voice disputes with logistics partners and platform support in Bahasa, Tagalog, or Thai. Platform support agents manually review audio clips — creating 72h+ resolution queues. Structured ASR at intake cuts resolution to same-day and enables automated routing.

Voice dispute clip→Structured dispute ticket + routed

Multi-language dispute entity extraction (order ID, amount, date, reason). Plugs into Zendesk / Freshdesk.

Agriculture & rural extension

Agriculture · Mekong

Cooperative harvest reporting hotline

Vietnamese coffee cooperatives in the Central Highlands and Cambodian rice cooperatives near the Mekong Delta run phone-in hotlines where farmers report harvest estimates, weather events, and pricing queries in rural Vietnamese and Khmer. Data is hand-typed by cooperative clerks 24–48 hours later, delaying commodity pricing and rural credit decisions.

Farmer IVR call→Harvest record + pricing system

Rural Vietnamese + Khmer + crop entity extraction. Bridges farmer to commodity market in real time.

Agriculture · CPEC corridor

Pakistan–China commodity price negotiation

Cotton, rice, and wheat traders in Punjab and Sindh negotiate prices with Chinese agri-buyers over WeChat voice notes entirely in Punjabi or Urdu. Chinese buyers respond in Mandarin. No ASR model covers this language pair. All terms are informally recollected, generating price disputes on multi-million-dollar seasonal commodity lots.

WeChat voice→Urdu / Punjabi ↔ Mandarin trade log

Urdu + Punjabi + Mandarin routing. First ASR solution for CPEC agri-trade. Zero incumbent coverage.

Construction & infrastructure

Construction · CPEC

CPEC site safety & daily progress reporting

Chinese construction managers at CPEC projects in Gwadar, Lahore, and Karachi require daily site safety and progress reports from Pakistani supervisors. Reports are submitted via WhatsApp voice in Urdu and Punjabi. Chinese project managers need Mandarin output for Beijing HQ. Currently translated manually with a 24-hour lag — safety incidents are documented after the fact.

Supervisor WhatsApp voice→Urdu → Mandarin site report

Real-time Urdu / Punjabi → Mandarin. First voice compliance layer on BRI’s $62B flagship corridor.

Construction · Mekong / SEA

Korean / Japanese ODA contractor ↔ Vietnamese subcontractor

JICA and KOICA-funded infrastructure projects across Vietnam and Cambodia operate with Korean and Japanese project managers alongside Vietnamese and Khmer subcontractors. Daily work instructions, handover reports, and materials requests travel by voice — no ASR coverage exists for Korean↔Vietnamese or Japanese↔Khmer. Safety and compliance records are reconstructed from memory.

Site radio / WhatsApp voice→Korean / Japanese ↔ Vietnamese structured log

Korean ↔ Vietnamese + Japanese ↔ Khmer routing. Covers $10B+ JICA / KOICA ODA portfolio in SEA.

Microfinance & financial inclusion

Microfinance · Bangladesh

BRAC / Grameen field officer loan notes

BRAC, ASA, and Grameen Bank field officers visit 20+ borrowers daily across rural Bangladesh, recording repayment status, income assessment, and household conditions via Bengali voice notes. Notes are manually transcribed at branch offices 24–48 hours later. Errors in borrower assessments affect credit models for 30M+ active loan accounts nationwide.

Field officer voice note→Structured loan record + CRM

Bengali ASR + microfinance entity extraction (borrower ID, repayment, income). 24h lag → real-time.

Financial inclusion · Philippines–Gulf

OFW remittance dispute resolution

2.2 million Filipino OFWs in Saudi Arabia, UAE, and Qatar dispute remittance transactions with local agents entirely over phone — Tagalog from the worker, Arabic from the agent. Disputes are unstructured voice calls that banks log manually. No structured speech capture exists for this $10B remittance corridor.

Dispute phone call→Tagalog ↔ Arabic dispute log

Tagalog + Gulf Arabic routing + financial entity extraction. Covers the world’s largest OFW remittance corridor.

Education & vocational training

Education · Myanmar / Thailand

Migrant worker vocational assessment

3 million+ Myanmar migrant workers in Thailand and Malaysia undergo vocational training assessments conducted in Thai or Bahasa, with learners responding in Burmese. No ASR model supports Burmese↔Thai or Burmese↔Bahasa. Assessments are scored via manual written translation — a 3–5 day delay between assessment and certification that blocks employment placement.

Assessment interview audio→Burmese ↔ Thai / Bahasa transcript + score

Burmese ASR — a language no incumbent covers. 3M+ underserved workers in a single corridor.

Education · Philippines

TESDA pre-deployment language assessment

TESDA conducts pre-deployment assessments for OFWs heading to Japan, Korea, and Gulf countries. Japanese and Korean proficiency must be verified through voice assessment — Tagalog-speaking candidates assessed in the destination country’s language. No speech assessment tool handles Tagalog↔Japanese or Tagalog↔Korean pairs. 500K+ assessments annually, all manually graded.

Assessment voice recording→Tagalog ↔ Japanese / Korean graded transcript

Tagalog + Japanese + Korean routing. 500K+ OFW pre-deployment assessments run annually in the Philippines.

Every use case runs on a single API integration.

One voice. Routed to the right engine.
Returned as structured data.

Talk to us about your use case →

Every voice,understood.

Global speech AI failswhere most of the world lives.

Orchestration, not invention.

Why VALSEA wins in the marketsAzure and Google ignore.

$6.9B market. Every enterprisethat touches non-English voice.

Enterprise proof pointsand pipeline.