A Berlin-based audio intelligence company founded in 2021 by Fabian Seipel and Corvin Jaedicke, with roughly $6.6M raised (a $1.6M pre-seed in 2024 and a $5M seed in 2025, backed by Connect Ventures, Partech, and Inovia Capital). Its positioning is audio reliability for Voice AI: when voice becomes core infrastructure, one misheard input costs trust. The product line splits by listener. Quail is optimised for speech-to-text and deliberately preserves some noise because machines process audio differently from humans—it claims up to 43% fewer detection errors than Krisp and SileroVAD with latency as low as 30ms. Rook is tuned for human listening in meetings and headsets. Voice Focus isolates the primary speaker and suppresses background voices for voice-agent turn-taking, and VAD ships as a standalone model for endpointing and STT cost savings. Tyto scores call audio quality and explains conditions likely to cause downstream failures. SDKs cover Python, Node.js, Rust, WebAssembly, C++, and C, running on-device so audio never leaves the device. A 30-day trial allows unlimited full-feature use including production, with no credit card; paid SDK subscriptions start at $149/month (up to $599/month) billed per audio minute with millisecond precision, and Enterprise adds on-prem deployment and custom audio evaluation. G2 shows 3.5/5 from 10 reviews, including two 2024 one-star reviews titled “service doesn’t work” and “doesn’t work, terrible support”, though the company replaced that older web app with a new developer platform in 2025.
**Pros & Cons**:
Pros:
Splits models by listener—Quail for STT accuracy, Rook for human listening—rather than assuming one enhancement profile fits both, which is the right architectural call for voice pipelines; claims up to 43% fewer STT detection errors versus Krisp and SileroVAD, with latency as low as 30ms; Voice Focus isolating the primary speaker addresses turn-taking in voice agents, a problem general noise suppression doesn’t solve; standalone VAD can cut STT costs by trimming silence; multi-language SDKs (Python, Node.js, Rust, WASM, C++, C) with on-device processing keep audio off external servers; EU-based infrastructure and a 30-day trial that permits production use lower adoption friction.
Cons: **The prior web app drew one-star G2 reviews** (“service doesn’t work”, “terrible support”) in 2024; the company sunset that app for a new developer platform in 2025, but the historical record should be noted; **pricing is confused across sources**—OMR lists €2–€30/month tiers that appear to belong to the old web app, while the current SDK subscription is $149–$599/month, so only the official pricing page should be cited; **no permanent free tier**—the 30-day trial ends and requires a paid plan or sales contact; **Quail deliberately preserves noise for STT**, so users seeking clean audio for human listening should use Rook, and the two should not be conflated; funding at roughly $6.6M is modest compared with larger infrastructure players, though the Berlin base and EU hosting are differentiators.
Who is it for?:Developers building voice agents or real-time transcription that need reliable audio input before the STT layer; teams whose pipelines break on noisy real-world audio and want enhancement tuned specifically for machines; voice-agent builders that need primary-speaker isolation and turn-taking support; organisations that require on-device processing and EU data residency.