MARS8 is a family of four specialised text-to-speech architectures designed for different production environments
CAMB.AI has launched MARS8, a text-to-speech (TTS) system designed as a family of specialised architectures built specifically for production.
For the last few years, CAMB.AI’s MARS engine has powered AI voice-enabled applications reaching over 200 million users through partners like NASCAR, Broadcom, Australian Open, Eurovision Sport, IMAX, and Comcast NBCUniversal.
After shipping voice AI into high-stakes production environments across 150+ languages, the team saw a gap in the market to create the MARS8 family of TTS models.
“The market forces developers to choose between speed, quality, accuracy, and cost. We realised that was a false choice,” says Prakash. “A live voice assistant needs sub-150ms latency. A movie dubbing pipeline needs director-level emotional control. An automotive system has strict memory constraints. You cannot solve all three with one generic API.”
MARS8 enables enterprises to run the right voice architecture for speed, quality, or efficiency at scale on their own infrastructure.
MARS8 offers four distinct architectures, each optimised for specific production constraints:
- MARS-Flash: Ultra-low latency, designed for real-time agents and call centers.
- MARS-Pro: The workhorse for expressive dubbing and digital media, balancing fidelity with speed.
- MARS-Instruct: Director-level control for high-end film production, allowing independent tuning of speaker and prosody.
- MARS-Nano: A highly efficient 50M parameter model for on-device applications where compute is constrained.
CAMB.AI says MARS8 “marks a turning point for production Voice AI. The days of forcing every use case through a single generic model are over. Whether you’re shipping a real-time voice agent, dubbing a feature film, or building an on-device assistant, you now have architecture built for your constraints, not compromises.”
No comments yet