Miso Labs
Miso Labs builds open-weights emotive text-to-speech voice models, launched from YC P26.
On the record
Checkpoint: Full-duplex speech-to-speech roadmap milestone, needed to match OpenAI's live voice product
The call
Miso Labs ships a fast open-weights voice model into a field where bigger, better-funded players already cover the same ground.
Miso Labs builds text-to-speech technology that turns written words into human-sounding voices. The problem is that Google and OpenAI already give away the same capability for free inside their products, and other well-funded startups like ElevenLabs are building nearly identical services. What could change things is if Miso finds one specific industry, like hospitals or banks that need to keep voice work private and local, and becomes the go-to choice there instead of trying to compete everywhere.
Why
01
AI pressure: Google Gemini TTS and OpenAI realtime voice. Google and OpenAI bundle comparable voice generation at near-zero marginal cost, squeezing standalone TTS pricing. Could AI do the core job? Yes, likely.
02
Bigger players: ElevenLabs. ElevenLabs, now valued at $11B after a $500M raise, runs the same open-weights emotive voice playbook with far more capital. Could a giant copy it? Yes, easily.
03
What protects it. Open weights let companies self-host for data control, a option the closed giants and ElevenLabs don't fully offer, though no paying regulated customer has been shown yet.
What would change this call
Miso ships a named regulated-industry, on-prem customer within 6-9 months. Would show Miso can own a defensible vertical niche instead of just competing on latency.
Miso announces full-duplex speech-to-speech capability matching OpenAI GPT-Live. Would show Miso closing the capability gap rather than permanently chasing parity.
Miso faces or avoids a Lovo-style voice-cloning lawsuit as NO FAKES Act advances. Would show whether legal exposure on voice cloning materializes into real liability.
Miso announces full-duplex speech-to-speech capability matching OpenAI GPT-Live. Would show Miso closing the capability gap rather than permanently chasing parity.
Miso faces or avoids a Lovo-style voice-cloning lawsuit as NO FAKES Act advances. Would show whether legal exposure on voice cloning materializes into real liability.
Key numbers
- 8B-parameter open-weights model launched June 3, 2026
- ElevenLabs valued at $11B after $500M raise
- Gradium raised $100M seed backed by Nvidia
Named risks
- Voice-cloning legal exposure amid tightening laws (NO FAKES Act, ELVIS Act) and Lovo's June 2026 bankruptcy precedent.
- No confirmed funding amount disclosed, only vendor-claimed 'large seed round' as of July 2026.
- Core latency/emotiveness features being commoditized by Google, OpenAI, and better-funded peers within months.
- Reputational damage from Tana Mongeau voice-mimicry controversy with no public company response.
- Small team (1-10 employees) attempting difficult full-duplex speech-to-speech pivot.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| ElevenLabs | losing to | ElevenLabs raised $500M at $11B valuation Feb 4, 2026, dwarfing Miso Labs' unconfirmed seed round |
| Sesame | losing to | Sesame runs the identical open-weight strategy with $250M raised (Oct 2025) versus Miso's undisclosed funding |
| Cartesia | losing to | Cartesia Sonic claims ~40ms latency versus Miso's claimed ~110ms as of May 2026 |
| Google (Gemini TTS) | losing to | Gemini 3.1 Flash TTS ties top of Speech Arena leaderboard with bundled zero marginal cost, mid-2026 |
| OpenAI | losing to | Acquired Weights.gg voice-cloning tech May 29, 2026, matching Miso's cloning feature at platform scale |
| Gradium | contesting | Closest peer in foundation-voice-model framing, raised $100M with NVIDIA backing by July 9, 2026 |
Questions people ask
- Is Miso Labs profitable?
- No revenue figures have been disclosed. Miso Labs is a pre-revenue, YC-backed (P26 batch) startup that launched its open-weight MisoTTS model on June 3, 2026, and no third-party or company revenue data exists as of July 10, 2026.
- Who owns Miso Labs?
- Miso Labs is privately held, founded 2025, part of Y Combinator's P26 (Spring 2026) batch. The company itself claims a large seed round from top investors, but no amount or lead investor has been publicly confirmed as of July 10, 2026.
- Could Google or OpenAI replace Miso Labs?
- Yes, largely. As of July 2, 2026, Google's Gemini 3.1 Flash TTS reportedly ties Cartesia Sonic 3.5 atop the Speech Arena leaderboard, and OpenAI acquired Weights.gg's voice cloning tech on May 29, 2026, both bundling comparable capability at zero marginal cost.
- Is Miso Labs a good company to join?
- It is a very early, 1-10 employee startup (as of July 2026) still hiring its Founding Research Scientist for $400K-$500K plus 2-3% equity, in a crowded, fast-commoditizing voice AI market with unconfirmed funding and a live voice-cloning legal controversy.
- Who is Miso Labs' biggest competitor?
- ElevenLabs is the closest incumbent threat, having raised $500M at an $11B valuation on February 4, 2026. Sesame, which raised $250M in October 2025, runs an identical open-weight strategy with far more capital than Miso Labs.
- Does Miso Labs face legal risk over voice cloning?
- Yes. Co-founder Aoden Teo publicly claimed the ability to clone any voice from 10 seconds of audio in June 2026, and creator Tana Mongeau flagged a Miso demo as mimicking her voice on June 11, 2026, amid tightening laws like the federal NO FAKES Act advanced June 18, 2026.
Timeline
- 2026-07-10Miso Labs posts Founding Research Scientist role at $400K-$500K plus 2-3% equity
- 2026-07-02Extruct.ai lists Miso Labs at 1-10 employees
- 2026-06-26Federal NO FAKES Act and state voice-cloning laws advance, tightening legal environment
- 2026-06-20Rival voice-cloning firm Lovo Inc. files Chapter 7 bankruptcy amid voice actor lawsuits
- 2026-06-11Creator Tana Mongeau flags Miso demo as mimicking her voice; Miso does not respond to press
- 2026-06-08MisoTTS GitHub repo confirms English-only, half-duplex model with future full-duplex plans
- 2026-06-03Miso Labs launches MisoTTS/Miso One, an 8B-parameter open-weights voice model
- 2026-05-29OpenAI announces acquisition of Weights.gg voice-cloning technology
- 2026-02-04ElevenLabs raises $500M from Sequoia at $11B valuation
- 2025-10Sesame raises $250M Series B co-led by Sequoia and Spark
Revision history
27 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- What the YC P26 batch tells you about where AI is going next (2026-06-18)
- Launch YC: Miso Labs - emotive voice models (2026-06-03)
- Open-Source Voice AI That Beats ElevenLabs on Latency (2026-06-05)
- Best Text-to-Speech Providers 2026 (2026-05-13)
- MisoLabsAI/MisoTTS · GitHub (2026-06-08)
- Open Weight 8B Voice Model with Voice Cloning (2026-07-01)
- Complete Analysis
- Founding Research Scientist at Miso Labs (2026-07-10)
- ElevenLabs raises $500M from Sequoia at an $11 billion valuation (2026-02-04)
- Paris-based AI voice startup Gradium raises $100M seed, backed by Nvidia (2026-07-09)
- Here are the 55 US AI startups that raised $100M or more in 2025 (2026-01-19)
- Deepgram raises $130M at $1.3B valuation and buys a YC AI startup (2026-01-13)