Real talk about AI voice tools — what they cost, how well they work, which ones speak Chinese
| Name↕ | Origin↕ | Freetier↕ | Pricing↕ | Voicequality↕ | Languages↕ | Cloning↕ | Api↕ | Realtime↕ |
|---|---|---|---|---|---|---|---|---|
| ElevenLabs 🇺🇸 USA | 🇺🇸 USA | 10K credits/mo | $6–$99/mo (Starter $6/30K, Creator $22/121K — $11 first month, Pro $99/600K); Scale $299/1.8M (3 seats), Business $990/6M (10 seats); annual ≈17% off | Studio (Eleven v3) | 70+ (v3) / 32 (Flash v2.5) including Chinese | Instant (30s) & Professional | Yes | Yes (Flash; v3 not realtime) |
| PlayHT (PlayAI) 🇺🇸 USA | 🇺🇸 USA | None — service shut down | DISCONTINUED. Meta acqui-hired the ~35-person PlayAI team in Jul 2025 into Superintelligence Labs; the API went dark Jul 26, 2025, signups stopped Aug 2025, and the platform was permanently terminated Dec 31, 2025. All accounts, voice clones and API keys deleted with no export path. play.ht no longer resolves. | — | — | — | No — shut down Dec 31, 2025 | No |
| Azure Speech 🇺🇸 USA | 🇺🇸 USA | 500K chars/mo | $16/1M (Neural), $22/1M (Neural HD — cut from $30 Mar 2026), custom tiers $24–$48/1M; commitment tiers from $7.50/1M | Neural / Neural HD / Custom Neural | 500+ voices, 140+ languages, Chinese | Professional (Custom Neural Voice) | Yes | Yes |
| Google TTS 🇺🇸 USA | 🇺🇸 USA | 4M standard chars/mo (WaveNet included); 1M each on Neural2/Chirp 3 HD/Studio | $4–$160/million chars (Standard/WaveNet $4; Neural2 $16; Chirp 3 HD $30; Instant Custom Voice $60; Studio $160). Gemini-TTS is token-priced instead | Studio (highest), Chirp 3 HD, Neural2, WaveNet, Standard | 220+ voices, 40+ languages, Chinese | Yes (Instant Custom Voice — 10s sample, $60/1M) | Yes | Yes |
| Amazon Polly 🇺🇸 USA | 🇺🇸 USA | Legacy: 5M Standard + 1M Neural chars/mo (1yr); new accounts: $200 credits | $4/1M (Standard), $16/1M (Neural), $30/1M (Generative), $100/1M (Long-Form) — global region rates | Standard / Neural / Generative (Sonantic) / Long-Form | 80+ voices, 60+ languages, Chinese | Limited (Personalize Voice beta) | Yes | Yes |
| Resemble AI 🇺🇸 USA | 🇺🇸 USA | Flex pay-as-you-go ($0 base, credits never expire) | Usage-based Flex + Business + Enterprise. In Jul 2026 Resemble REMOVED all voice-generation per-second rates from its public pricing page — it now publishes deepfake detection ($0.03–$0.07/sec), identity search ($0.0005/search) and watermarking (encode $0.0005/sec, decode $0.0002/sec) with voice moved under "Voice AI Research". Voice cloning: first clone free, then $2 each; team seats $20/user/mo. Enterprise is a custom quote | Neural / Studio | 10+ including Chinese (cloned) | Instant (30s) & Professional | Yes | Yes |
| Speechify 🇺🇸 USA | 🇺🇸 USA | Limited free reader | Premium $29/mo or $139/yr (~$11.58/mo — the annual plan is the whole point); Studio $19–$49/mo (credit-based, separate product); Audiobooks ~$9.99–$14.99/mo separate | Neural / Studio (Pro) | 30+ including Chinese | Instant (Pro tier) | Yes (SpeechifyAI API, $6–$10/1M chars) | Yes |
| Lovo (Genny) 🇺🇸 USA | 🇺🇸 USA | None — company bankrupt | SHUT DOWN. Lovo, Inc. filed Chapter 7 liquidation on May 27, 2026 (S.D.N.Y., case 1:26-bk-11249) — liquidation, not reorganisation — while the voice-actor lawsuit (Lehrman v. Lovo) was stayed Jun 11, 2026. Paid users reported lockouts from late Apr 2026; genny.lovo.ai is offline. Do not start anything new here. | — | — | — | No — company in Chapter 7 | No |
| Descript 🇺🇸 USA | 🇺🇸 USA | 60 media min/mo (720p watermarked exports) | $16–$65/mo (Hobbyist $24, Creator $35, Business $65 monthly; $16/$24/$50 annual — annual saves up to ~35%) | Good (Overdub) | English only (Overdub) | Instant (Overdub) | Yes | No |
| iFlytek 🇨🇳 China | 🇨🇳 China | Limited daily calls | Pay-as-you-go / Enterprise | Best-in-class Chinese | Primarily Chinese, English, Japanese, Korean | Professional & Instant | Yes | Yes |
| OpenAI TTS 🇺🇸 USA | 🇺🇸 USA | $5 new-account credits | $15/1M chars (tts-1), $30/1M (tts-1-hd); gpt-4o-mini-tts is token-priced at $0.60/1M text input + $12/1M audio output (~$0.015/min) | Good (standard) / HD | 6 voices, English + multi-language | No | Yes | Yes |
| Murf AI 🇺🇸 USA | 🇺🇸 USA | Free: 10 min generation, 1 editor, no downloads and no commercial rights | Creator $19/mo annual ($29 monthly, 2 hrs/mo); Business $66/mo annual ($99 monthly, 8 hrs/mo); Enterprise custom. API billed separately from $0.01/1K chars (Falcon) and $0.03/1K (Studio-quality). Unused minutes do not roll over | Studio (200+ voices, 20+ styles) | 20+ languages including Mandarin | Limited — check current plan docs | Yes (pay-as-you-go per character) | Limited |
| Cartesia (Sonic) 🇺🇸 USA | 🇺🇸 USA | Free: 20K credits/mo | From $5/mo (Sonic-3.6, ~133 min included, 3 concurrent requests); approx 1 credit per character for TTS, 1.5 credits/char for Pro Voice Cloning after a one-time training fee; Managed Agents $0.06/min + $0.014/min telephony; enterprise custom | Sonic-3.6 (sub-100ms latency, built for realtime) | Multi-language (Sonic voice library) | Instant & Professional (training fee applies) | Yes | Yes |
| Deepgram 🇺🇸 USA | 🇺🇸 USA | $200 free credits (no expiry) | Aura-2 $30/1M chars ($27/1M on Growth); Flux TTS $0.045/1K chars PAYG, $0.0405 on Growth — the free Flux launch promo ended Sep 12, 2026; Voice Agent API ~$4.50/hr. Deepgram is matching Flux spend with credits up to $500 through Dec 31, 2026 | Good neural TTS (Aura-2, Flux for agents) | 7 languages (Aura-2): English, Spanish, French, German, Dutch, Italian, Japanese | No (Aura-2) | Yes | Yes |