How accurate is AssemblyAI speech-to-text on drug names?

Medical Mode cuts entity errors by about 87% versus the base model, and medications are a core entity class — including sound-alike pairs like hydroxyzine and hydralazine.

Can I bias transcription toward my pharmacy formulary?

Yes — pass your drug list in keyterms_prompt (up to 1,000 terms async).

Does it capture dosages and frequencies?

Yes — numbers, units (mg, mL), and frequency shorthand (BID, q.d.) are part of the entity coverage.

Does it work in real time or batch for pharmacy calls?

Both — realtime for live calls, async for recorded interactions.

Is patient data protected and BAA-eligible?

AssemblyAI is a business associate under HIPAA and signs a standard BAA; PHI redaction; SOC 2 Type 2.

How do I enable Medical Mode for pharmacy workflows?

Add domain: "medical-v1" — $0.15/hr, no model change.

Pharmacy

When the drug name is the whole point

In pharmacy workflows, a transcription error isn't a typo — it's the wrong medication. Medical Mode is tuned for exactly this: drug names, dosages, and sound-alike differences (hydroxyzine vs. hydralazine). 3.2% Missed Entity Rate, one parameter, async or realtime.

Try Medical Mode free on your audio Read the docs

87% fewer entity errors than the base model — and medications are entities

Missed Entity Rate 3.2%

Missed Entity Rate on clinical conversations with Medical Mode.

vs. base model ~20%

Fewer missed medical entities than the base model without Medical Mode.

vs. base model 87%

Fewer medical entity errors than the base model.

Benchmarked providers #1

Lowest Missed Entity Rate vs. Deepgram, Speechmatics, AWS, and Google.

See the medical benchmarks

Activation

Bias toward your formulary

Pass your formulary as keyterms to bias toward the drugs you actually dispense — one parameter, no model swap.

import assemblyai as aai

aai.settings.api_key = "YOUR_API_KEY"

# Pharmacy: sound-alike drug names, dosages
config = aai.TranscriptionConfig(
    speech_models=["universal-3-5-pro", "universal-2"],
    domain="medical-v1",
    keyterms_prompt=["hydroxyzine", "hydralazine", "metformin", "10 mg"],
)

transcript = aai.Transcriber().transcribe("./pharmacy-call.wav", config)
print(transcript.text)

Sound-alike drug names

Reduces the errors that turn one medication into another.

Universal-3.5 Pro

Dosage and frequency

Numbers, units, and frequency shorthand captured precisely.

Medical Mode

Formulary keyterms

Pass your formulary as keyterms to bias toward the drugs you actually dispense.

Universal-3.6 Pro Realtime

Common questions

: Medical Mode cuts entity errors by about 87% versus the base model, and medications are a core entity class — including sound-alike pairs like hydroxyzine and hydralazine.
: Yes — pass your drug list in keyterms_prompt (up to 1,000 terms async).
: Yes — numbers, units (mg, mL), and frequency shorthand (BID, q.d.) are part of the entity coverage.
: Both — realtime for live calls, async for recorded interactions.
: AssemblyAI is a business associate under HIPAA and signs a standard BAA; PHI redaction; SOC 2 Type 2.
: Add domain: "medical-v1" — $0.15/hr, no model change.

Try Medical Mode free on your audio

Run your own pharmacy audio through Medical Mode in the playground — one parameter, no model swap. Free to start, no credit card required.

Try Medical Mode free