← all releases
CAMB.AI·Jan 20, 2026·v1.0·paid

MARS8

MARS8 is a family of specialized text-to-speech (TTS) models designed for real-world deployment, enabling enterprises to choose the optimal voice architecture for speed, quality, or efficiency at scale on their own infrastructure. The MARS8 family includes four distinct architectures: MARS-Flash, MARS-Pro, MARS-Instruct, and MARS-Nano, each optimized for specific production constraints.

AudioTTS
visit website
overview

MARS8 is a family of specialized text-to-speech (TTS) models designed for real-world deployment, enabling enterprises to choose the optimal voice architecture for speed, quality, or efficiency at scale on their own infrastructure. The MARS8 family includes four distinct architectures: MARS-Flash, MARS-Pro, MARS-Instruct, and MARS-Nano, each optimized for specific production constraints.

key features
  • 01MARS-Flash: Ultra-low latency (sub-150ms TTFB) designed for real-time agents and call centers, optimized for Blackwell GPUs, L4 and L40S.
  • 02MARS-Pro: Designed for expressive dubbing and digital media, balancing fidelity with speed.
  • 03MARS-Instruct: Provides director-level control for high-end film production, allowing independent tuning of speaker and prosody.
  • 04MARS-Nano: A highly efficient 50M parameter model for on-device applications where compute is constrained.