Scribe V2 vs. AssemblyAI

Learn why developers choose AssemblyAI Universal-3.5 Pro over ElevenLabs Scribe V2 to build production Voice AI:

Lower English WER on real-world audio (4.35% vs 5.87%)
Streaming transcription with speaker diarization—Scribe V2 Realtime does not
Simpler pricing with no per-feature add-ons, and a BAA that isn’t Enterprise-gated

Get your API key See the comparison

Universal-3.5 Pro

Your transcriptions will show here...

At a glance: Scribe V2 vs. AssemblyAI

Model

AssemblyAI Universal-3.5 Pro

ElevenLabs Scribe V2

English WER on real-world audio

4.35%

5.87%

Code-switching WER (5-language avg)

7.60%

10.07%

Speaker diarization (aggregate cpWER)

35.21%

Published as % accuracy

Streaming diarization

—

Pre-recorded pricing

$0.21 / hour

$0.22 / hour

Real-time pricing

From $0.15 / hour

$0.39 / hour

Add-on fees for entities & keyterms

Metered separately

Free tier

$50 in free credits

10,000 credits (no card)

BAA available

Enterprise + ZRM only

PCI DSS

—

Go beyond transcription with Assembly's full Voice AI Infrastructure

Best-in-Class Accuracy

Universal-3.5 Pro Async posts a 4.35% English WER on real-world audio—beating Scribe V2 at 5.87%—and stays ahead on multilingual and code-switched audio.

Streaming + Diarization

Get ultra-low-latency streaming transcription with speaker labels in the same stream. Scribe V2 Realtime does not support diarization.

Speaker Diarization

Universal-3.5 Pro leads on speaker diarization with 35.21% aggregate cpWER—a full-word-level benchmark, not a self-reported label-match percentage.

Contextual Prompting

Guide transcription with domain context and vocabulary. In healthcare tests, contextual prompting cut missed medical terms by 31%.

Native Code-Switching

Native code-switching across 18 languages, with automatic fallback to Universal-2 for 99 languages of coverage overall.

Simple Pricing, No Add-ons

Pay $0.21/hr batch and from $0.15/hr real-time—no separate charges for entity detection or keyterm prompting like Scribe V2 charges on top.

PII Redaction & BAA

Detect and redact PII from transcripts and audio, and sign a BAA in minutes for apps that process PHI—not gated behind an Enterprise tier.

Proven Reliability and Security

Deploy on infrastructure that processes millions of hours daily, with 99.9% uptime, unlimited concurrency, and SOC 2 Type 2, ISO 27001, PCI DSS, and GDPR.

Start building

Get your free API key and ship your first transcript in minutes—no commitments or minimums.

Ready to outgrow Scribe V2?

Switch to higher accuracy, streaming with diarization, and pricing without per-feature add-ons—with no contracts or minimums.

Get your API key

Investments in STT improvements always pay for themselves, since it is such a critical building block of the voice pipeline.

Lindsay Liu, Co-Founder & CEO at Super

Playground

We're not playing around—but you can

Test our best-in-class speech-to-text and voice agent models in our no-code playground.

Explore Playground

Frequently asked questions

: AssemblyAI is more accurate in AssemblyAI’s benchmarks: Universal-3.5 Pro posts a 4.35% English WER on real-world audio versus 5.87% for Scribe V2, and it leads on code-switching (7.60% vs 10.07% average across five languages) and speaker diarization (35.21% cpWER, an industry-standard word-level metric Scribe V2 does not publish). Scribe V2 posts strong results on FLEURS and Common Voice—clean, read-aloud corpora—but AssemblyAI benchmarks on the messy audio customers actually send.
: Yes, on comparable workloads. AssemblyAI is $0.21/hour pre-recorded and from $0.15/hour real-time. Scribe V2 batch starts at $0.22/hour but climbs to ~$0.34/hour once you turn on entity detection (+$0.07/hr) and keyterm prompting (+$0.05/hr). Scribe V2 Realtime is $0.39/hour—roughly 2.6× AssemblyAI real-time.
: No. Scribe V2 Realtime does not support speaker diarization—ElevenLabs states this is not a priority for the real-time model. AssemblyAI supports streaming transcription with speaker labels in the same stream, at around 300ms latency, which is critical for live meeting bots and agent-assist workflows.
: Scribe V2 claims 90+ languages, with real-time accuracy published on 30 of them. AssemblyAI supports 99 languages, with Universal-3.5 Pro delivering native code-switching across 18 of them and automatic fallback to Universal-2 for the rest. If broad multilingual coverage is a priority, evaluate both on your own audio.
: Yes. AssemblyAI signs a Business Associate Addendum (BAA) in minutes without a sales call, at every tier—not just Enterprise. ElevenLabs offers HIPAA support only on its Enterprise tier and requires Zero Retention Mode. AssemblyAI is also certified for SOC 2 Type 2, ISO 27001, PCI DSS, and GDPR.
: Yes. The AssemblyAI API is a small delta from Scribe V2 for most workloads, and SDKs are available in Python and JavaScript/TypeScript. Sign up, get $50 in free credits, and swap the endpoint to run both models side by side on your own audio.