Forty 55-minute earnings calls transcribed: MAI Streaming vs Sume STT

Transcribing the earnings calls of forty companies in one quarter is 2,200 minutes of audio. That is 2,200 audio minutes, about $19.80 on MAI-Transcribe-2-Streaming ($0.54 an audio hour, an introductory rate Microsoft says runs through the end of 2026) and $22.40 on Sume STT 1.0 ($0.01 per audio minute) including the one-cent split job that cuts each recording into pieces of 10 minutes or less.

Inputs are assumptions: forty 55-minute English calls in a quarter. Microsoft's rate is from its announcement and Sume's from the API reference, pricing page and timeline audio docs, all read 2026-10-05.

What the bill looks like

Each call is six transcription jobs on Sume (five of 10 minutes and one of 5), so a quarter is 240 jobs plus 40 splits.

Names and numbers decide whether the transcript is usable

Calls are full of tickers, product names and figures. Microsoft's MAI-Transcribe-2 page lists keyword biasing for domain terminology. Sume STT 1.0 has no keyword list on the request: it takes audio_url, language_code, duration_seconds, optional segmentation and metadata. Sume STT 1.0 takes duration_seconds from 1 to 600, so a recording longer than 10 minutes is cut into pieces first. The timeline audio split job takes up to 20 ranges, so one split covers up to 200 minutes of audio.

• Correct company names in a second pass, using the words[] timings to find them.
• Pass metadata such as the ticker and quarter; it is stored with the job request and not sent to the provider.
• Use a stable idempotency key per call and chunk, such as call-q3-ACME-2, so a rerun is safe.
• The recordings must be reachable at a public HTTPS URL; Sume prefers its own media host.

Run it on Sume

This transcribes each call from a list of hosted audio URLs and writes one text file per call.

The recording URL passed in must already be a media.sume.com audio file, because the split job only reads Sume-hosted audio; import it first.

import os, time, requests
API = "https://api.sume.com"
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
def run(path, body, key):
    r = requests.post(API + path, headers={**H, "Idempotency-Key": key}, json=body, timeout=60)
    r.raise_for_status()
    job = r.json()["data"]
    while True:
        s = requests.get(job["status_url"], headers=H, timeout=30).json()["data"]
        if s["terminal"]:
            break
        time.sleep(2)
    if s["sume_status"] != "completed":
        raise RuntimeError(s["sume_status"])
    return requests.get(job["result_url"], headers=H, timeout=30).json()["data"]["result"]
def transcribe_long(media_url, total_s, key):
    ranges = [{"start": a, "end": min(a + 600, total_s)} for a in range(0, total_s, 600)]
    split = {"operation": "split", "url": media_url, "ranges": ranges, "output": {"format": "mp3"}}
    parts = run("/v1/timeline-1.0/audio", split, key + "-split")["segments"]
    texts = []
    for i, p in enumerate(parts):
        body = {"audio_url": p["audio_url"], "language_code": "en",
                "duration_seconds": min(600, total_s - 600 * i)}
        texts.append(run("/v1/stt-1.0/transcribe", body, f"{key}-{i}")["text"])
    return " ".join(texts)

calls = os.environ["CALL_URLS"].split(",")
for i, url in enumerate(calls, 1):
    text = transcribe_long(url, 55 * 60, f"call-q3-{i:03d}")
    open(f"call-{i:03d}.txt", "w", encoding="utf-8").write(text)

When MAI is the better pick

For finance vocabulary, keyword biasing is the feature to weigh: it is listed on the MAI-Transcribe-2 page and it is not in the Sume request. The price gap of $2.60 per quarter is small next to that difference.