Two-hour webinar transcript: Sume STT $1.20 in 12 jobs vs other rates

A 2-hour webinar costs $1.20 to transcribe on Sume STT: 12 jobs of 10 minutes at $0.10 each. Published rates from three other vendors for the same two hours range from about $0.30 to $0.90, whether the audio is streamed live or sent as a file. Sume is the dearest per minute of the group and is file-only; it earns its place only if the transcript feeds the rest of your Sume pipeline.

Vendor rates are from the pricing pages of Deepgram, AssemblyAI and ElevenLabs, read 2026-10-03. Two-hour totals are my arithmetic.

Streaming and file rates for two hours

Streaming means words arrive while the speaker talks, which suits live captions. File means you send the recording afterwards. A recorded webinar only needs the second.

Why twelve jobs

Sume STT takes one public HTTPS audio URL per job and caps duration at 600 seconds. Two hours is 7,200 seconds, so 12 parts. Cut the audio at fixed 600-second marks, or at silences just before them, upload each part, and submit the jobs. Word timings are always returned, so each part's times are offset by where the part started. The stitching is covered in splitting a long recording.

A planner you can run

The function below returns the part count and cost in cents for a recording length in minutes, and the start offset of every part, using the 10-minute cap and the $0.01 a minute price. It does no network calls.

import math

CAP_MIN = 10        # 600 seconds per STT job
CENTS_PER_MIN = 1   # $0.01 a minute

def plan(total_min):
    parts = math.ceil(total_min / CAP_MIN)
    offsets = [i * CAP_MIN * 60 for i in range(parts)]
    return parts, total_min * CENTS_PER_MIN, offsets

parts, cents, offsets = plan(120)
print(parts, 'jobs, $%.2f' % (cents / 100))
print('part start offsets (s):', offsets)

A weekly webinar for a year

Fifty-two one-hour sessions is 52 hours. Sume STT is $0.60 an hour, so $31.20 a year in 312 jobs. ElevenLabs Scribe v2 at $0.22 an hour is $11.44, and AssemblyAI pre-recorded at $0.15 to $0.21 an hour is $7.80 to $10.92. The gap is about $20 a year in cash. What it buys is one integration and a transcript already in the form the clip, caption and render steps expect, so judge it by the engineering time you save.

What Sume does and does not do

Sume does not stream audio in or captions out. It returns a finished transcript for each job. Once you have it, the same API can pick moments, trim, caption and render clips; see webinar recording to shorts.