By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Silicon Valley Girl

$6.6B AI CEO: How to Make Your First $10,000 with AI

Get the best domain for your business with or use coupon code: Marina to get your perfect domain for just 99 cents for the first year.

Mati StaniszewskiguestMarina Mogilkohost

Oct 4, 202543mWatch on YouTube ↗

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

ElevenLabs’ voice marketplace and the “$10k/month with voice agents” hook
Marina introduces Mati Staniszewski, CEO/co-founder of ElevenLabs, framing the company’s scale and the core promise: voice AI is now realistic enough to create new income streams. The episode tees up two parallel threads—monetizing voice (marketplace/royalties) and monetizing deployments (voice agents for businesses).
- •ElevenLabs positioned as a $6.6B leader in Voice AI
- •Voice marketplace enables voice cloning and passive income
- •Immediate opportunity teased: deploying voice agents to businesses for ~$10k/month
- •Central tension introduced: powerful voice tech vs misuse/control
Why voice becomes the primary AI interface (beyond text)
Mati argues voice will be a key interface for AI because it carries richer information than text. He highlights emotional nuance, inflection, and a more natural user experience as drivers of adoption.
- •Voice conveys emotionality, imperfections, and context better than text
- •Better input understanding and more pleasurable output experience
- •Rapid shift since early ChatGPT era from text-first to multimodal/voice-first
Voice agents in the funnel: support, product guidance, inbound/outbound sales
They discuss how voice agents replace legacy IVR and improve customer support, then extend into the full user journey. Mati shares that ElevenLabs uses agents to answer product/pricing questions, route leads, and sometimes convert customers directly.
- •Voice agents outperform old IVR: faster, more natural, higher satisfaction
- •Use cases expand from support into onboarding/product navigation
- •ElevenLabs uses agents for faster pipeline movement and lead qualification
- •Direct conversion possible for self-serve tiers; enterprise still needs KYC
Sponsor break: securing a brandable domain with .online
Marina pauses the interview to discuss the difficulty of finding good .com domains and promotes .online as an alternative. The pitch emphasizes availability, brand fit, and a limited-time discount offer.
- •Problem: short, brand-matching .com domains are often taken
- •Solution: .online domain extension with broad adoption
- •Examples of .online usage and SEO/discoverability claims
- •Call to action: discount link and coupon code
How to set up an ElevenLabs voice agent: orchestration, business logic, workflows
Mati explains the mechanics of deploying voice agents on ElevenLabs’ platform. The system abstracts technical complexity (speech + LLM + TTS) while the business must supply knowledge bases, rules, and workflows (e.g., appointment scheduling).
- •Agent platform bundles low-latency speech/LLM/TTS orchestration
- •Business must provide knowledge base and desired Q&A materials
- •Workflow logic: triggers, function calls, and predefined templates
- •Example: scheduling flow that checks calendars and confirms slots
Selling courses via AI calls—multilingual, omnichannel, and website embed
Marina maps the concept to her course business and explores how agents can sell in multiple languages using her voice. Mati describes purchase flows (links, email follow-up, embedded web agent) and how the agent can guide users through checkout.
- •Agents can speak multiple languages while preserving the creator’s voice identity
- •Omnichannel flow: call → link/email → checkout subscription
- •Website embed: agent guides users through forms in real time
- •Potential to lower anxiety for non-native speakers (AI ‘doesn’t judge’)
Cost and integrations: telephony (Twilio), existing numbers, and SMB pricing
They get practical about what it costs and what’s required to connect calls. Mati estimates hundreds of dollars per month to start (volume-dependent) and notes integration with telephony providers like Twilio using existing phone numbers.
- •Starting cost estimate: hundreds of dollars/month depending on call volume
- •Telephony integration via Twilio/other systems
- •Bring your existing phone number; IP calling supported
- •Channel strategy: using voice AI to reduce language barriers vs WhatsApp-only
Getting paid for your voice: marketplace mechanics, royalties, and what earns more
Mati explains how creators authenticate, record ~30 minutes, and produce a voice that can be shared to the marketplace under defined conditions. He shares payout totals and discusses why uniqueness (accent, prosody, style) drives higher earnings.
- •Voice creation requires authentication + ~30 minutes of recordings
- •Voices can be shared to marketplace; creators earn when others use them
- •Community payouts grew from ~$2M to ~$5M+ (approaching ~$10M over time)
- •Average varies; many creators earn ‘a few hundred/month’ with some promotion
- •Unique voices/accents can outperform common-sounding voices
Voice cloning quality issues: matching intonation, scene context, and audio mixing
Marina describes a real editing workflow—patching new phrases into existing videos—and why regenerated audio can sound slightly off. Mati outlines why “average voice” training misses scene-specific emotion/intonation and suggests short-term and future fixes.
- •Mismatch comes from scene-level intonation/emotion vs averaged voice model
- •Background sound/processing differences can make inserts feel unnatural
- •Future idea: condition on a few seconds before/after the edit point for better morphing
- •Short-term workaround: regenerate multiple times or clone from a shorter, scene-specific sample
Where voice AI is heading: personal authenticated agents and voice personalization
They project a near future where everyone has an AI voice clone and a personal agent that can act on their behalf. Mati predicts businesses will tailor voices to customer segments and users will choose preferred voices for services and navigation.
- •Rise of ‘your’ authenticated voice agent that can perform tasks (booking, follow-ups)
- •Voice verification methods will need to evolve beyond simple voice auth
- •Businesses can serve different voices/styles for different demographics or regions
- •User-preference voices could become standard (e.g., navigation narration going viral)
Deepfakes and safeguards: the three-layer model (device, watermark, default distrust)
Marina presses on impersonation risk when voice is tied to identity and payments. Mati lays out a three-layer approach: prove human via trusted device, watermark authenticated AI, and otherwise assume audio is AI and treat it as untrusted.
- •Assumption shift: perfect voice clones will exist; systems must be designed accordingly
- •Layer 1: verify ‘human’ via trusted/registered devices and signals
- •Layer 2: watermark authenticated AI (with quality trade-offs in audio)
- •Layer 3: if not verified/watermarked, treat content as AI by default and don’t trust it
- •Operational safeguards: traceability to accounts, moderation, scam-text detection
Founder anxieties: staying ahead in research, responsibility, and economic disruption
Mati shares what keeps him up at night: maintaining a research edge, building safety systems, and managing labor-market impacts responsibly. He frames AI as an opportunity bigger than the internet, but one that requires careful stewardship.
- •Pressure to keep innovating across TTS, STT, orchestration, and music
- •Responsibility for misuse prevention and safety investments
- •Concern about job disruption and designing inclusive participation
- •Belief that it’s still early; hiring and culture are critical
Jobs at risk and adaptation: domain expertise + AI, and shifting support work
They discuss which roles are most exposed (repetitive support tasks) and how people can adapt. Mati emphasizes that people will be replaced by people using AI, and that domain expertise becomes more valuable as routine work is automated.
- •Repetitive, recipe-based tasks (appointments/refunds) are easiest to automate
- •Complex edge cases still require human experts—at least for now
- •Core advice: use AI tools to stay at the frontier; become ‘AI-augmented’
- •Combine domain expertise with AI to increase leverage and job resilience
- •AI accelerates iteration cycles; economy can expand rather than be zero-sum
Top AI tools + the $10k/month SMB voice-agent opportunity and startup advice
Mati names favorite tools (besides ElevenLabs) and then spells out the immediate business opportunity: deploying voice agents for small local businesses that don’t know it’s possible. He closes with entrepreneurship guidance—obsess over a real problem, validate demand, and choose co-founders carefully—plus reflections on language learning’s future.
- •Tool picks: Black Forest Labs (images), Claude, Lovable (plus v0/Replit/Figma)
- •$10k/month path: deploy voice agents to SMBs (dentists, mechanics) for scheduling/intake
- •No-code/low-code angle: you don’t need to be an engineer; you need outreach + setup
- •ElevenLabs origin story: Polish ‘single-voice’ dubbing pain → pivot to creator post-production and audiobooks
- •Startup advice: problem obsession, validate what users actually want, pick co-founders/early team well
- •Languages may shift from necessity to hobby/culture, with translation devices reducing friction

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.