AssemblyAI Universal-3.5 Pro $0.21 per hour vs Sume STT

AssemblyAI's pricing page lists Universal-3.5 Pro for asynchronous transcription at $0.21 per hour, Universal-2 at $0.15 per hour, and a speaker-diarization add-on at $0.02 per hour. Sume STT 1.0 is $0.01 per audio minute ($0.60 per hour). At list price AssemblyAI is roughly a third of Sume's rate.

What the page lists

AssemblyAI also lists a realtime model and $50 in free credits. The table covers only the async lines.

Sume job example

Sume's transcript comes back as text plus words[] with start and end times.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/stt-1.0/transcribe",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
             "Idempotency-Key": "stt-demo-001"},
    json={"audio_url": "https://example.com/call.mp3",
          "language_code": "en",
          "duration_seconds": 300},
)
print(r.status_code, r.json())

Where this leaves you

If transcription cost is your largest line and you do not need other media steps, AssemblyAI's published rate is lower. If the transcript is an input to captions, dubbing or timeline audio, staying inside one job model can save integration work. Measure on your own audio before deciding.

Worked example

Speaker labels change the total, so here is the full line.

• 100 hours of Universal-3.5 Pro: $21.00; with speaker labels at +$0.02 per hour: $23.00.
• 100 hours of Universal-2: $15.00.
• Sume STT for 100 hours: $60.00.

Checklist before you commit

Sume's docs say diarization and event tags are fixed server-side, so there is no separate switch to price. Check a sample result to see how speakers appear in your output.

• Decide whether you need labelled speakers at all.
• Check AssemblyAI's streaming billing if you will also stream: it is per session duration.
• Compare accuracy on noisy audio from your own use case.