Any language. Any voice. Instantly.

Real-time multilingual speech AI for everyday use

Step 1 — Paste the reference text you expect to say

Type or paste the exact phrase you will read aloud for this test (speak it verbatim for accurate WER). Example: "Hi, my name is John Doe and my phone is 555-1234."
Tip: include a short sentence (5–20 words). This demo may include example PII for testing — avoid using real sensitive information if you prefer.

WER% (Word Error Rate) for transcript

WER: 0%(0% is best, 100% is worst)

Number of words: 0 |Substitutions: 0 |Deletions: 0 |Insertions: 0 |Correct: 0

0–10% Excellent10–20% Good20–30% Fair30–50% Poor>50% Very poor

View alignment

Op	Reference text	Transcript text

Step 2 — Verbum redaction & profanity filter

Configure how Verbum should redact sensitive content and how profanity should be handled. These settings will be applied when the socket is created before recording.

Redaction categories

Profanity filter

Step 3 — Select the language and record (speak for ~5–10s)

1) Choose the language you'll speak. 2) Click “Start Recording”, say the reference text clearly for about 5–10 seconds, then click “Stop Recording”.

Verbum:

Verbum Transcription

Deepgram Transcription

Verbum vs Deepgram

Verbum

Server location:

Deepgram

Accuracy

Latency

-ms

Accuracy [%] (Higher is better). Latency [ms] (Lower is better).

Feature comparison — Verbum vs Deepgram

Feature	Verbum	Deepgram
Speech-to-Text	Multilingual (82 languages, 40 dialects)	Multilingual (dozens supported)
Text-to-Text Translation	Yes, multilingual	Not native, workflow-based
Text-to-Speech	Yes, multilingual	Yes, via API/stream
Real-Time Streaming	Yes, near real-time	Yes, real-time
Transcription Accuracy	Multilingual, noisy conditions	High accuracy, diarization available
Speaker Diarization	Yes, in-call/multi-user	Yes, multiple models
Security Encrypted data in transit	Enterprise-grade compliance (SOC2, HIPAA, GDPR)	SOC-2, HIPAA, PCI DSS, and GDPR
Custom Vocabulary / Models	Yes, custom vocab/models	Yes, custom vocab/models
Chat-Auto Translation	Yes, bidirectional	No, requires external translation
Management APIs	Yes, project/key management	Yes, project/user/key management
Use Cases	Meetings, calls, chat, captions	Contact center, meetings, analytics
Customization & Integration	REST and Websocket	SDKs, deep integration, REST/ws

Speech-to-Text

Verbum

Multilingual (82 languages, 40 dialects)

Deepgram

Multilingual (dozens supported)

Text-to-Text Translation

Verbum

Yes, multilingual

Deepgram

Not native, workflow-based

Text-to-Speech

Verbum

Yes, multilingual

Deepgram

Yes, via API/stream

Real-Time Streaming

Verbum

Yes, near real-time

Deepgram

Yes, real-time

Transcription Accuracy

Verbum

Multilingual, noisy conditions

Deepgram

High accuracy, diarization available

Speaker Diarization

Verbum

Yes, in-call/multi-user

Deepgram

Yes, multiple models

Security Encrypted data in transit

Verbum

Enterprise-grade compliance (SOC2, HIPAA, GDPR)

Deepgram

SOC-2, HIPAA, PCI DSS, and GDPR

Custom Vocabulary / Models

Verbum

Yes, custom vocab/models

Deepgram

Yes, custom vocab/models

Chat-Auto Translation

Verbum

Yes, bidirectional

Deepgram

No, requires external translation

Management APIs

Verbum

Yes, project/key management

Deepgram

Yes, project/user/key management

Use Cases

Verbum

Meetings, calls, chat, captions

Deepgram

Contact center, meetings, analytics

Customization & Integration

Verbum

REST and Websocket

Deepgram

SDKs, deep integration, REST/ws