← All market maps
Voice & Audio AI Market Map: TTS, Real-Time Voice Agents, Music & Video Dubbing
A comprehensive interactive directory of 90 core Voice & Audio AI players across the US/Global, Korea (KR), Japan (JP), China (CN), and Europe (EU). Spanning the entire stack: ① Emotional TTS & Zero-Shot Voice Cloning (ElevenLabs, OpenAI Audio, Cartesia, Fish Audio, Supertone, Typecast, Qwen3-TTS) → ② Real-Time Voice Agents & Ultra-Fast STT Infrastructure (Deepgram, AssemblyAI, LiveKit, Vapi, Retell AI, Bland AI, Return Zero, CLOVA Speech) → ③ AI Music Generation, Vocal Synthesizers & SFX (Suno, Udio, Stability Audio, Pozalabs, VOCALOID6, Synthesizer V, Mureka, ACE-Step) → ④ Multilingual Video Dubbing & Lip-Sync Localization (HeyGen, Rask AI, ESTsoft Perso.ai, Papago Dubbing, Sync Labs, CAMB.AI). Filter and sort instantly by Scale (Valuation/Market Cap), Usage, Region (US/Global, KR, JP, CN, EU), Category (cat 1–4), Tier (Tier 1, Tier 2, Rising, Open Source), and 3-tier Verification Badge (✔ Verified / △ Estimated / ✘ Unconfirmed).
Key Numbers Defining the Industry
300ms · 75ms300ms is the human conversational turn-taking threshold — Cartesia Sonic (40–90ms TTFB), ElevenLabs Flash (75ms), and OpenAI Realtime S2S shattered the latency barrier for natural voice agents · ✔
$3.3B · $200M+ElevenLabs reached a $3.3B Series C valuation and ~$200M+ estimated ARR — vertically integrating TTS, dubbing, SFX, and conversational agents · ✔
$0.05 – $0.15/minTotal all-in cost per minute for an AI phone agent (STT + LLM + TTS + WebRTC/SIP) — cutting human call center labor costs ($1.00–$2.50/min) by 85–95% · △
12M+ · $500MSuno surpassed 12M+ creators at a $500M valuation ($125M Series B) — driving mainstream AI songwriting alongside open-source full-song models (ACE-Step, YuE) despite RIAA lawsuits · ✔
90% Saved · 175 LangsAI video dubbing & lip-sync (HeyGen, Rask AI, Perso.ai, Sync Labs) slashed localization costs from $50–$150/min in traditional studios down to $1–$3/min · ✔
99 Players · 4 Pillars① TTS & Voice Cloning (39) · ② Real-Time Voice Agents & STT (25) · ③ AI Music & Sound Effects (19) · ④ Video Dubbing & Localization (16) across US, KR, JP, CN, EU & Open Source · ✔
How the Market Map Is Structured — 4 Core Categories × 5 Regions × 4 Tiers × 3-Tier Verification
The Voice & Audio AI ecosystem has evolved far beyond robotic screen readers into a four-layer value chain: synthesizing & cloning expressive human voices (① TTS & Voice Cloning) → listening, reasoning, and handling live phone/WebRTC calls in <300ms (② Real-Time Voice Agents & STT) → composing full vocal songs and royalty-free sound effects (③ AI Music & SFX) → translating videos and re-rendering speaker lip movements across 175+ languages (④ Video Dubbing & Localization). Every player is classified by headquarters region (US/Global, KR, JP, CN, EU), market tier (Tier 1, Tier 2, Rising, Open Source), and a 3-tier verification badge (✔ Official/Verified, △ Estimated/Reported, ✘ Unconfirmed/Rumor).
① TTS & Voice CloningElevenLabs · OpenAI Audio · Cartesia · PlayHT · Fish Audio · Supertone · Typecast · Qwen3-TTS · CosyVoice 2 · GPT-SoVITS · Hume AI · Murf · Rime · CLOVA Voice · VOICEVOX · MiniMax Audio · Seed-TTS · Kokoro-82M. Zero-shot voice cloning from 3–10s audio clips and sub-100ms streaming synthesis.
② Real-Time Voice Agents & STTDeepgram · AssemblyAI · LiveKit Agents · Vapi · Retell AI · Bland AI · OpenAI/Groq Whisper · Gemini Live · Gladia · Speechmatics · Pipecat · Return Zero · CLOVA Speech · Daglo · PKSHA · AmiVoice · Kotoba · iFlytek · SenseVoice · Kyutai Moshi. Barge-in handling, SIP telephony, and sub-200ms ASR.
③ AI Music & Sound EffectsSuno · Udio · Stability Audio · ElevenLabs SFX · Google Lyria (MusicFX) · Meta MusicGen · SOUNDRAW · AIVA · Boomy · Pozalabs · Neutune (MixAudio) · Mubert · Yamaha VOCALOID6 · Synthesizer V · Kunlun Mureka · ByteDance Sponge Music · NetEase Tianyin · ACE-Step/YuE.
④ Video Dubbing & LocalizationHeyGen Video Translate · Rask AI · ESTsoft Perso.ai · Naver Papago Dubbing · ElevenLabs Dubbing Studio · Sync Labs · CAMB.AI · Papercup · Deepdub · Dubverse · Hudson AI · Voithru · Twigfarm · CapCut Dubbing · Panjaya · Sieve. Multi-speaker isolation, emotional translation & 4K lip-sync.
Filter & Verification Badge (✔ / △ / ✘) Guide· Region (region): Filter across 5 geographic hubs:
US/Global, KR (South Korea), JP (Japan), CN (China), and EU (UK, France, Germany, Luxembourg, Sweden, etc.).· Tier (tier): Tier 1 (Unicorns, Big Tech & market leaders), Tier 2 (Established enterprise & vertical leaders), Rising (Fast-growing breakout startups), and Open Source (Open-weight models & frameworks).· 3-Tier Verification Badge (v): ✔ Officially verified via filings, funding releases, pricing pages, or GitHub repos; △ Estimated or media-reported private metrics; ✘ Unconfirmed rumors or unreleased standalone products.Player Directory
Region
Category
Tier
Verification
Domain
Capabilities
Architecture
Revenue Model
Ownership
Sort
| # | Name | Region | Category | Tier | Primary Use Cases | Key Metric & Valuation | Revenue Model | Verify |
|---|
Click any row to expand headquarters, ownership, latency benchmarks, pricing, copyright licensing, recent trends, and primary source (src) links.
Market Size, Latency & Unit Economics Benchmarks
| Segment / Metric | Value / Range | Definition & Methodology (2025–2026) | Verify |
|---|---|---|---|
| Global Generative Voice & Audio AI Market | $6.5B – $9.2B (2026E) · 34%+ CAGR | Combined revenue across cloud TTS APIs, real-time voice agent infrastructure, AI music/SFX generators, and video dubbing SaaS | △ |
| Real-Time Voice Agents & AI Contact Center (AICC) | $2.4B – $3.8B (2026E) · $0.05–$0.15/min all-in cost | Deepgram, LiveKit, Vapi, Retell AI, Bland AI & enterprise AICC deployments replacing legacy IVR trees and human BPO seats | △ |
| Ultra-Low-Latency Streaming TTS (TTFB) | 40ms – 90ms (SSM / Flash models) | Official benchmarks from Cartesia Sonic (40–90ms), ElevenLabs Flash v2.5 (75ms), and Rime Mist (<100ms), enabling <300ms round-trip voice agents | ✔ |
| High-Speed STT Transcription Pricing & Accuracy | $0.0007 – $0.0077/min · ~5% WER | Groq Whisper Turbo ($0.04/hr), Deepgram Nova-3 ($0.0043/min), AssemblyAI Universal-2, and Return Zero VITO official pricing | ✔ |
| AI Music Generation Volume & SaaS Pricing | Tens of millions of tracks/mo · $10–$30/mo Pro tiers | Suno (12M+ creators, $500M valuation ✔), Udio, SOUNDRAW & Mureka pricing — ~$0.02–$0.05 per full 4-minute vocal song | ✔ |
| AI Video Dubbing & Lip-Sync Cost Reduction | $1.00 – $3.50/min (90%+ cheaper than studios) | HeyGen, Rask AI, Perso.ai, Sync Labs & ElevenLabs Dubbing Studio vs. $50–$150/min for traditional human studio dubbing | ✔ |
Business Models — 6 Ways Money Flows in Voice & Audio AI
| Revenue Model | Representative Players & Pricing | Structural Dynamics & Margin Trends |
|---|---|---|
| ① Character & Credit-Metered TTS SaaS / API | ElevenLabs ($5–$330/mo · $0.05–$0.18/1K chars) · OpenAI Audio ($15/1M chars) · Typecast · Supertone Play | While basic TTS per-character pricing faces downward pressure from open-source models (Kokoro, Fish Speech, Qwen3-TTS), leaders defend ARPU via voice cloning slots, emotion control, and studio workflows. |
| ② Per-Minute Voice Agent Orchestration | Vapi · Retell AI · Bland AI · Synthflow ($0.05–$0.09/min platform fee + STT/LLM/TTS/SIP pass-through = $0.09–$0.18/min total) | Abstracts away complex WebRTC jitter buffers, barge-in detection, and SIP trunking. Benchmarked directly against human call center wages ($1–$2/min), commanding strong enterprise willingness to pay. |
| ③ Native Speech-to-Speech (S2S) Audio Tokens | OpenAI GPT-4o Realtime API ($40/1M input · $80/1M output audio tokens; Mini at 1/4 cost) · Google Gemini 2.0 Live API | Collapses STT→LLM→TTS into a single multimodal transformer preserving laughter and tone. Delivers unmatched conversational fluidity, though long calls require VAD and context caching to control token burn. |
| ④ Prosumer AI Music Subscriptions vs. Copyright-Safe B2B | Suno · Udio · Mureka ($10–$30/mo) ↔ Stability Audio · SOUNDRAW · Pozalabs · AIVA (indemnified B2B licensing) | Suno and Udio dominate consumer viral songwriting, while enterprise game studios, broadcasters, and ad agencies sign B2B contracts with 100% licensed/in-house trained models to avoid RIAA litigation risk. |
| ⑤ Per-Minute Video Dubbing & 4K Lip-Sync Credits | HeyGen ($29–$39/mo) · Rask AI ($60–$750/mo) · ESTsoft Perso.ai · Sync Labs ($1.50–$3.00/min) · Sieve | Visual lip-sync re-rendering commands 3–5x higher per-minute margins than audio-only dubbing, though CapCut Pro ($10–$20/mo) is aggressively bundling translation for short-form creators. |
| ⑥ Voice Actor IP Marketplaces & Royalty Sharing | ElevenLabs Voice Library · Supertone (HYBE artist IP) · CoeFont · Synthesizer V / VOCALOID Voicebanks · Respeecher | Compensates voice actors, singers, and celebrities with per-character royalties or upfront licensing fees whenever their cloned voice is used commercially. |
Architecture Comparison — 3 Stacks for Building Real-Time Voice Agents
Comparing end-to-end latency, per-minute unit economics, and production trade-offs across the main voice agent stacks in 2026.
| Architecture Stack | Typical Component Combination | E2E Latency | Est. Cost / Min | Key Strengths & Production Trade-offs |
|---|---|---|---|---|
| ① Modular Pipeline (Best-of-Breed STT + LLM + TTS) | LiveKit Agents / Pipecat + Deepgram Nova-3 + GPT-4o-mini/Claude + Cartesia Sonic / ElevenLabs Flash | 350ms – 600ms | $0.04 – $0.08/min | #1 enterprise production choice — offers full deterministic control over text transcripts, tool calling, guardrails, and vendor swapping |
| ② Managed Voice Agent Platform (Turnkey API) | Retell AI / Vapi / Bland AI (bundles top STT, LLM, TTS, barge-in VAD & Twilio/SIP telephony) | 400ms – 700ms | $0.09 – $0.18/min | Fastest time-to-market — solo builders and startups can deploy inbound/outbound AI phone agents with CRM webhooks in hours |
| ③ Native Speech-to-Speech (Single S2S Model) | OpenAI GPT-4o Realtime API / Google Gemini 2.0 Live / Kyutai Moshi (Open Source) | 200ms – 350ms | $0.06 – $0.30+/min | Zero intermediate text bottleneck — hears and responds to emotion, whispers, and interruptions natively, though pricier on long calls |
| ④ Self-Hosted Open Source (Private GPU Cluster) | LiveKit + Faster-Whisper / SenseVoice + vLLM (Llama/Qwen) + Fish Speech / Kokoro-82M / Orpheus | 400ms – 800ms (GPU dependent) | $0.01 – $0.02/min (compute) | Essential for air-gapped finance, healthcare, defense, and high-volume contact centers optimizing marginal compute cost |
Major Deals, Funding & Open-Source Milestones (2024–2026)
| Date | Player & Milestone | Scale & Strategic Significance | Verify |
|---|---|---|---|
| 2022-01 | HYBE Acquires Controlling Stake in Supertone | KRW 45B (~$32M+) investment — brought expressive singing & multilingual voice cloning in-house for K-Pop IP | ✔ |
| 2024-05 | Suno Raises $125M Series B | Valued at $500M led by Lightspeed & Nat Friedman — catalyzed mainstream consumer AI songwriting | ✔ |
| 2024-06 | RIAA Sues Suno & Udio for Copyright Infringement | Universal, Sony, and Warner Music filed landmark lawsuits seeking up to $150,000 per infringed work | ✔ |
| 2024-06 | HeyGen Raises $60M Series A at $500M Valuation | Led by Benchmark — viral breakout of 175-language video translation and photorealistic lip-sync | ✔ |
| 2024-07 | Kyutai Open-Sources 'Moshi' Full-Duplex S2S Model | European AI lab released open weights for a 200ms native speech-to-speech conversational model | ✔ |
| 2024-10 | OpenAI Launches Realtime API & Advanced Voice | Opened native GPT-4o audio-in/audio-out streaming to developers globally | ✔ |
| 2024-12 | Cartesia Raises $64M Series A | Backed by Index & Lightspeed — scaled Mamba SSM-based Sonic TTS with 40–90ms latency | ✔ |
| 2025-01 | ElevenLabs Raises $180M Series C at $3.3B Valuation | Co-led by a16z & ICONIQ — expanded from TTS leader into full Conversational AI agent orchestration | ✔ |
| 2025-02+ | Open-Source Wave: Qwen3-TTS, CosyVoice 2, Kokoro-82M & ACE-Step | Alibaba, StepFun, and open-source researchers released Apache 2.0 TTS and full-song music models rivaling closed SaaS | ✔ |
7 Key Trends Reshaping Voice & Audio AI (2025–2026)
- From Flat 'Reading' to Emotional 'Acting' & Sub-300ms 'Conversing': Basic informational TTS has been commoditized by ultra-cheap APIs and lightweight open-source models (Kokoro-82M), shifting premium value toward dramatic emotional prosody (Hume, ElevenLabs, Supertone, Hudson AI) and sub-300ms real-time turn-taking (Cartesia, OpenAI Realtime).
- Modular Pipelines (STT+LLM+TTS) Coexist with Native Speech-to-Speech (S2S): While GPT-4o Realtime and Gemini Live set the bar for natural cadence, enterprise contact centers still heavily favor modular pipelines (Deepgram + LLM + Cartesia/ElevenLabs orchestrated via LiveKit, Vapi, or Retell) for strict transcript auditability, deterministic tool use, and lower cost.
- State Space Models (SSMs) Conquer Low-Latency Audio Streaming: Pioneered by Cartesia (founded by Stanford Mamba creators), SSM architectures deliver constant memory footprint and 40–90ms time-to-first-audio, outperforming autoregressive transformers in live telephony.
- The Great Bifurcation in AI Music — Viral Prosumer Hits vs. Indemnified B2B: Amid ongoing RIAA litigation against Suno and Udio, the market split into two tracks: prosumer vocal song generators (Suno, Udio, Mureka) vs. 100% licensed/in-house trained copyright-safe platforms (Stability Audio, SOUNDRAW, Pozalabs) favored by studios and brands.
- Open-Source Audio Models Close the Gap with Proprietary SaaS: Alibaba's CosyVoice 2 & Qwen3-TTS, Fish Speech, GPT-SoVITS, Kokoro-82M, and full-song model ACE-Step allow developers to self-host zero-shot voice cloning and music synthesis under Apache 2.0 / MIT licenses.
- Video Dubbing Evolves into Full-Body Visual Localization: Audio-only dubbing is giving way to 4K zero-shot lip-sync (HeyGen, Sync Labs, Perso.ai), full-body gesture retiming (Panjaya BodyTalk), and live sports crowd-preserved commentary dubbing (CAMB.AI).
- Mandatory Audio Watermarking & Deepfake Defense: Imperceptible neural watermarks (Google SynthID, Meta AudioSeal) and real-time deepfake detectors (Resemble Detect) have become mandatory enterprise compliance guardrails against voice phishing and unauthorized cloning.
Six Standout Hubs & Players by Region
US / Global — ElevenLabs & Cartesia
- ElevenLabs ($3.3B valuation): The undisputed full-stack audio AI leader spanning TTS, dubbing, SFX & conversational agents
- Cartesia (Sonic): Redefined real-time voice agent responsiveness with 40–90ms Mamba SSM streaming synthesis
Also watch: OpenAI Audio · Deepgram · LiveKit · Suno · HeyGen
South Korea (KR) — Supertone · Typecast · Return Zero · Perso.ai
- Supertone (HYBE) & Typecast (Neosapience): Global pioneers in emotional character acting TTS and singing voice synthesis
- Return Zero (VITO) & Naver CLOVA: Dominating Korean enterprise AICC, meeting transcription, and Papago Dubbing
- ESTsoft (Perso.ai) & Pozalabs: Leading K-content lip-sync video localization and 100% copyright-clean AI music
Also watch: NC AI VARCO Voice · ActionPower (Daglo) · Hudson AI · Voithru
Japan (JP) — VOICEVOX · Synthesizer V · PKSHA · SOUNDRAW
- VOICEVOX & CoeFont: Powering Japan's massive Zundamon creator wave and rights-cleared voice actor marketplace
- Synthesizer V (Dreamtonics) & Yamaha VOCALOID6: Studio-grade AI vocal synthesizers dominating J-Pop & virtual singer production
- PKSHA & AmiVoice & Kotoba: Enterprise call-center speech AI and low-latency Japanese speech-to-speech
Also watch: SOUNDRAW (ethical global AI beats) · CeVIO AI
China (CN) — Fish Audio · CosyVoice 2 · MiniMax · CapCut
- Alibaba (CosyVoice 2 / SenseVoice / Qwen) & GPT-SoVITS: The backbone of the global open-source voice cloning and ASR community
- Fish Audio & MiniMax Audio: Top-ranked multilingual zero-shot TTS engines with massive global developer adoption
- CapCut (ByteDance) & Mureka (Kunlun): 500M+ MAU editor-bundled dubbing and full-song DiT music synthesis
Also watch: iFlytek · ACE-Step / YuE · IndexTTS (Bilibili)
Europe (EU) — Kyutai · Gladia · Speechmatics · Stability Audio
- Kyutai (France): Proved European open-science leadership with the 200ms full-duplex Moshi speech-to-speech model
- Speechmatics (UK) & Gladia (France): Enterprise accuracy leaders across 50–100+ languages and code-switched audio
- Stability Audio (UK), Papercup (UK) & AIVA (Luxembourg): Copyright-safe SFX/music, broadcast HITL dubbing & MIDI composition
Also watch: Synthflow AI (Germany) · Soniox · Acapela
Open Source — Whisper · LiveKit · Pipecat · Kokoro-82M
- OpenAI Whisper & SenseVoice: The foundational open-weight ASR models powering global transcription pipelines
- LiveKit Agents & Pipecat (Daily): The two default open-source WebRTC frameworks for real-time voice agents
- Kokoro-82M, F5-TTS & Orpheus: Proved that sub-100M to 3B open models can rival commercial TTS APIs on a single GPU
Also watch: GPT-SoVITS · Meta AudioCraft · Dia-1.6B
Changes
- First public edition (99 players across 4 categories in US/Global, KR, JP, CN, and EU · 3-tier verification badges ✔/△/✘ with primary source links · per-minute economics, sub-300ms latency stack, open-weight licenses & ToS guardrails · new Toki studio cover across 4 languages).
※ Disclosure: Some tool links are partner affiliate links. We may earn a commission at no extra cost to you.
Company information
- Company
- Abover Co., Ltd.
- Address
- 302-1128 (Sejong Plaza), 25 Jangcha-ro 13beon-gil, Gochon-eup, Gimpo-si, Gyeonggi-do, Republic of Korea
- Business registration no.
- 361-86-03930
- Terms · Privacy
- Terms of service · Privacy policy
© 2026 ainewground · Published by Abover Co., Ltd.