We do sales calls and we are currently using Cartesia. What do you think are some very good, available, natural human-like voice models other than Cartesia and ElevenLabs?

Thanks for reaching out, Waleed. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

Ny other than these ?

Thanks — great question. Besides Cartesia, ElevenLabs, Inworld, and Rime, other natural, human-like TTS options you can try (all supported via LiveKit either through Inference or plugins) are: • Deepgram Aura-2 — professional, expressive voices with prosody controls. • xAI Text to Speech — energetic, confident voices (very low cost). • Google Cloud TTS — wide language/voice selection, SSML and voice-cloning options. • Azure Speech TTS — high-fidelity neural/generative voices with SSML control. • Amazon Polly — neural & generative engines with SSML for fine prosody. • Plus plugin-supported providers: Groq, Mistral AI, Sarvam, Hume, Resemble AI, Neuphonic, Speechify, Speechmatics, Spitch, Smallest AI. If you want, I can: • Recommend 2–3 specific voices to test based on your buyer profile (region, age, tone). • Provide a short script and settings (pitch/rate/SSML hints) to audition them. Which audience or tone should I tailor suggestions to? Sources: Text-to-speech (TTS) models overview | LiveKit Documentation | Deepgram TTS | LiveKit Documentation | Pricing | LiveKit | Google Cloud TTS integration guide | LiveKit Documentation | Azure Speech TTS plugin guide | LiveKit Documentation, with interactive elements

They all seems to be robotic

Thanks — I hear you: sounding natural (not robotic) is critical for sales calls. To help quickly, a couple of directions that get markedly more human-like: • Use a custom or fine-tuned voice (voice cloning / custom voice) rather than an out-of-the-box voice — this preserves natural idiosyncrasies and cadence. Providers with strong custom voice capabilities: Resemble AI, Azure Speech (Custom Neural/Voice Font), Google Cloud (custom voice / Chirp). • Enable and tweak prosody/SSML features (pauses, emphasis, pitch, speaking-rate, and breaths). Small SSML changes often transform a “robotic” voice into a convincing conversational one. • Pick voices designed for conversational/real-time use (phones/live agents) rather than broadcast/read-aloud voices; Inworld Realtime and Deepgram Aura-2 emphasize conversational patterns. • Add utterance variation: alternate short/long sentences, insert natural filler (brief “mm-hm”, “okay”), and use dynamic phrasing so replies don’t sound identical every time. Quick questions to narrow recommendations: 1. Do you need live/real-time TTS or pre-recorded outbound calls? 2. Which region/accents work best for your prospects? 3. Do you want to clone a specific salesperson’s voice or use a generic natural voice? Tell me those and I’ll recommend 2–3 concrete provider+voice choices and a short SSML demo script to test. , with interactive elements

Which of these beats elevenlabs

Good question — thanks for clarifying. If your top priority is sounding more natural than ElevenLabs, these options tend to outperform it for conversational, sales-style speech (based on voice naturalness, customization, and prosody control): • Resemble AI — Strong custom/voice-cloning capabilities that preserve speaker idiosyncrasies; very effective for making a voice sound genuinely human. • Azure Speech (Custom Neural / Voice Font) — High-fidelity neural voices with strong SSML and custom voice options for realistic prosody. • Google Cloud TTS (custom voice / Chirp) — Wide range of high-quality voices and voice-cloning that can create very natural conversational tones. • Deepgram Aura-2 — Designed for expressive, conversational speech with prosody controls that reduce “robotic” feel. • Inworld Realtime TTS — Built for conversational applications and real-time responsiveness, which helps naturalness in live calls. Why these beat ElevenLabs for sales calls: • Custom/voice-clone options (Resemble, Azure, Google) keep natural idiosyncrasies and cadence. • SSML/prosody controls and real-time conversational models (Azure, Google, Deepgram, Inworld) let you add pauses, emphasis, breaths, and variations that remove a robotic monotone. If you want, I can next: • Recommend 2 specific providers/voices to audition for your buyer region and tone, and give a short SSML demo script to test naturalness; or • Help you evaluate cloning one of your top sales reps’ voices vs. using a top-tier generic voice. Which would you prefer — a cloned rep voice or a generic natural voice for testing? , with interactive elements

esemble AI — Strong custom/voice-cloning capabilities that preserve speaker idiosyncrasies; very effective for making a voice sound genuinely human. what pricing for ths

I don’t have Resemble AI’s current pricing in the sources. Resemble is used via the LiveKit plugin (you’ll need your own Resemble API key), and pricing varies by features (cloning vs. standard voices), usage (minutes/characters), and enterprise options. If you want, I can: • Pull Resemble’s public pricing page and summarize it for you, or • Estimate costs for your usage if you tell me expected minutes per month and whether you need voice cloning. Which would you prefer? Sources: Text-to-speech (TTS) models overview | LiveKit Documentation | Resemble AI TTS plugin guide | LiveKit Documentation | Build Your First AI Voice Agent in Python: Complete Tutorial | LiveKit | Prompting guide | LiveKit Documentation | A Practical Guide to Prompting Gemini 3.1 Flash TTS | LiveKit, with interactive elements