Speech recognition built for domain precision.
One unified developer API for General, Financial, Medical, and Real-Time Streaming audio. Drop-in OpenAI SDK compatible at up to 73% lower cost.
See domain conditioning in action
Compare how Verba's vertical acoustic-linguistic models capture spoken stock tickers, executive names, clinical dosages, and live voice turns without post-processing hacks.
Four specialized models. One API key.
Switch domains or move from batch files to live WebSockets by changing a single
model parameter.
Verba Speech General
Built for meetings, podcasts, media archives, customer support calls, and accented global English in noisy environments with zero silence hallucinations.
Verba Speech Finance
Trained for corporate earnings calls, analyst Q&A, expert network interviews, spoken stock tickers, basis points, and GAAP/non-GAAP financial figures.
Verba Speech Medical
Precision clinical transcription for ambient AI scribes, telehealth, pharmacology regimens (mg, mL, PO/BID), lab panels, and stateless zero-data-retention audio.
Verba Speech Realtime
Sub-240ms streaming speech-to-text with neural voice activity detection (VAD) and
turn endpointing. Supports general, finance, and
medical modes live.
Accuracy that outperforms legacy clouds at a fraction of the price
Word Error Rate (WER — lower is better) across industry-standard held-out benchmarks.
| Benchmark / Metric | Verba Speech v1 | Deepgram Nova-3 | AssemblyAI Universal-2 | OpenAI Transcribe |
|---|---|---|---|---|
| SPGISpeech (Corporate Earnings Calls WER) | 2.38% | 2.95% | 3.10% | 3.50% |
| LibriSpeech Clean (General English WER) | 1.48% | 1.83% | 1.60% | 1.90% |
| LibriSpeech Other (Noisy & Accented WER) | 2.66% | 3.90% | 3.10% | 3.60% |
| PriMock57 Medical (Clinical Term Recall) | 98.2% | 96.8% | 96.5% | 96.4% |
| Pay-As-You-Go Price (Per Audio Hour) | $0.10 / hr | $0.26 – $0.45 / hr | $0.37 / hr | $0.36 / hr |
Drop-in compatible in 30 seconds
Point the standard OpenAI SDK at https://verbaspeech.com/v1 or open a
standard WebSocket for real-time voice agents.
Start with 5 free hours every month. Forever.
No expiring 14-day trials, no vertical surcharges for Medical or Finance, and per-second billing when you scale.
- ✓ 5 hours / month free forever
- ✓ Access to General, Finance, Medical & Realtime
- ✓ Custom vocabulary biasing (
prompt) - ✓ Word-level timestamps & confidence
- ✓ No credit card required
- ✓ $0.00167 / minute across all 4 models
- ✓ High-concurrency batch & WebSocket streaming
- ✓ Stateless Zero Data Retention mode included
- ✓ Prepaid credits never expire
- ✓ Instant API key scaling
- ✓ Committed-volume discounts
- ✓ Signed HIPAA BAA & security review
- ✓ Dedicated reserved streaming concurrency
- ✓ Custom institutional / clinical lexicons
- ✓ Priority engineering Slack channel