Soniox TTS v2 Launches at Just $0.70 Per Generated Hour
rohanpaul_ai · x · 2026-08-12
Soniox has introduced Soniox TTS v2, a new text-to-speech model delivering premium voice quality and extremely low latency at a breakthrough price of $0.70 per generated hour.
- All-in-One: Combines high-fidelity voice cloning, 60+ languages, natural language mixing, and low-latency streaming in a single model.
- Real-time Agents: Built specifically for live AI agents with character-level timestamps, allowing agents to start talking early, stop cleanly when interrupted, and resume seamlessly.
- New Architecture: Features a newly designed model architecture, audio codec, and optimized inference engine, replacing complex traditional voice-stack plumbing.
More from Multimodal
- DMSampler Accelerates Diffusion RL Training, Cutting GPU Hours by 10x — jiqizhixin · 2026-08-12
- Minimax H3 vs. Flux 3: Image Generation Showdown — Devajyoti1231 · 2026-08-12
- LTX 2.5 Video Model Disappoints: Users Report No Visible Improvement — PuppetHere · 2026-08-12
- Lightx2v vs. Larryvrh Turbo: Controlled Video Generation Test — SaskinPikachu · 2026-08-12
- Gaussian Splats Are Here: Turning the Real World into a Directable Scene — umesh_ai · 2026-08-12
- People Are Rebooting 'The Twilight Zone' Using AI Video Tools — astralmatrix · 2026-08-12