Deepgram, a voice AI platform, on Tuesday launched Aura-2, its next-generation text-to-speech (TTS) mannequin. The corporate calls it the world’s {most professional} and cost-effective enterprise-grade TTS answer.
In blind assessments by customers particularly for conversational enterprise functions, the mannequin outperformed main opponents like ElevenLabs, Cartesia, and OpenAI.
Aura-2 is constructed on prime of Deepgram Enterprise Runtime (DER), a customized infrastructure layer for its speech fashions. It goals to supply domain-specific pronunciation, skilled voice high quality, and context-aware supply with the speech generated.
With this, builders can improve real-time enterprise interactions throughout varied use circumstances, together with customer support, digital brokers, and AI-powered assistants.
Aura-2 may be deployed by way of cloud or on-premises APIs. Furthermore, new customers will obtain $200 in free credit to strive the mannequin’s capabilities on the official web site.
The corporate explains a major hole in enterprise-optimised voice AI, which requires a natural-sounding voice and domain-specific pronunciation. Deepgram’s Aura-2 makes an attempt to bridge this hole for business-critical environments.
“In head-to-head comparisons throughout enterprise eventualities, Deepgram got here out on prime practically 60% of the time,” the corporate acknowledged. As per the chart shared, Aura-2 was most well-liked by customers 61.8% in comparison with 38.2% for ElevenLabs. Equally, a choice of 52% may be seen compared to 48% for OpenAI.
When requested concerning the mannequin’s totally different use circumstances, Natalie Rutgers, VP of product for Deepgram, instructed AIM: “Whereas folks can use Aura-2 for podcasts and different leisure use circumstances, that isn’t our focus with this providing. Our prospects care about having real-time voices that characterize the folks you’d hear at your appointments, your pharmacy, and your customer support traces.”
Rutgers additionally talked about that the mannequin helps English voices, together with British and Australian accents, with multilingual help underway.
Deepgram’s Aura-2 can be optimised for real-time efficiency. It claims to ship quick response instances, with a sub-150ms time-to-first-byte.
The mannequin claims to supply the bottom pricing in comparison with ElevenLabs Flash and Cartesia Sonic. Deepgram explains, “At $0.030 per 1,000 characters, it gives substantial financial savings in comparison with alternate options like Elevenlabs Turbo ($0.050) and Cartesia Sonic ($0.038).”
The corporate states that usage-based pricing eliminates high quality/price tradeoffs, enabling uniform voice experiences at each touchpoint whereas sustaining efficiency and managing prices.
The publish Deepgram’s New Textual content-to-Speech AI Mannequin Outperforms ElevenLabs and Open AI appeared first on Analytics India Journal.
