Real-time Benchmarks

Measured on live audio streams to quantify WER, background speaker suppression, and VAD stability under real-world conditions.

Lower WER in real time

Up to 10-25% relative WER reduction across major ASR providers.

Stable turn detection

Higher F1 score and balanced accuracty than Silero VAD.

Fewer background insertions

Cleaner input means higher ASR accuracy, smarter VAD, and steadier LLM.

Quail

Benchmarked against leading STT models and VAD solutions — delivering lower error rates across every test.

Word error rates on commercial STT models

Raw

Quail Voice Focus

Top part: Substitutions / Middle: Insertions / Bottom: Deletions

lower % is better

WERs

Across Deepgram, Soniox, AssemblyAl, Cartesia, and ElevenLabs, Quail Voice Focus cuts insertions by up to ~12%, improving transcription reliability in the presence of competing background speakers.

WERs

Across Deepgram, Soniox, AssemblyAl, Cartesia, and ElevenLabs, Quail Voice Focus cuts insertions by up to ~12%, improving transcription reliability in the presence of competing background speakers.

WERs

Across Deepgram, Soniox, AssemblyAl, Cartesia, and ElevenLabs, Quail Voice Focus cuts insertions by up to ~12%, improving transcription reliability in the presence of competing background speakers.

Download as image

Word error rates on commercial STT models

English subset

Raw

Krisp

Quail Voice Focus

Top part: Substitutions / Middle: Insertions / Bottom: Deletions

lower % is better

WERs on English subset

Quail achieves up to ~20% relative WER reduction across major STT models compared to raw audio and perceptual denoisers like Krisp.

WERs on English subset

Quail achieves up to ~20% relative WER reduction across major STT models compared to raw audio and perceptual denoisers like Krisp.

WERs on English subset

Quail achieves up to ~20% relative WER reduction across major STT models compared to raw audio and perceptual denoisers like Krisp.

Download as image

Word error rates on commercial STT models

German subset

Raw

Krisp

Quail Voice Focus

Top part: Substitutions / Middle: Insertions / Bottom: Deletions

lower % is better

WERs on German subset

The gains increase on German: Quail achieves up to ~25% relative WER reduction across the same STT models, where Krisp's perceptual denoising introduces more substitutions and deletions.

WERs on German subset

The gains increase on German: Quail achieves up to ~25% relative WER reduction across the same STT models, where Krisp's perceptual denoising introduces more substitutions and deletions.

WERs on German subset

The gains increase on German: Quail achieves up to ~25% relative WER reduction across the same STT models, where Krisp's perceptual denoising introduces more substitutions and deletions.

Download as image

VAD Performance Comparison

Silero VAD

Quail VAD

lower % is better

Outperforms Silero VAD

Quail VAD outperforms Silero on every metric — F1 Score 0.767 vs. 0.724, Balanced Accuracy 0.777 vs. 0.760. More accurate speech detection, fewer missed segments, better results.

Outperforms Silero VAD

Quail VAD outperforms Silero on every metric — F1 Score 0.767 vs. 0.724, Balanced Accuracy 0.777 vs. 0.760. More accurate speech detection, fewer missed segments, better results.

Outperforms Silero VAD

Quail VAD outperforms Silero on every metric — F1 Score 0.767 vs. 0.724, Balanced Accuracy 0.777 vs. 0.760. More accurate speech detection, fewer missed segments, better results.

Download as image