Transcribe 500 hours: Speechmatics tiers vs Sume STT cost

Transcribing a 500-hour archive costs $60 on Speechmatics Melia 1, $115 on its Standard tier, $190 on Enhanced and $300 on Sume STT 1.0, using the rates on each page read on 2026-10-05. Speechmatics is cheaper on every tier it lists, and Melia 1 is a fifth of the Sume price. The comparison is price only, and it ignores accuracy, languages and diarization.

The rates

Speechmatics lists Melia 1 at $0.12 per hour, Standard at $0.23 per hour and Enhanced at $0.38 per hour, with add-ons priced separately (translation, for one, is listed at $0.49 per hour). Sume STT 1.0 is $0.01 per audio minute, which is the provider list of $0.008 times 1.25, or $0.60 per hour. The table multiplies each by 500 hours.

Why Sume is higher

Sume bills provider list times 1.25 in integer USD micros, and the STT model behind it is billed at $0.008 per minute before that margin. The gap to Speechmatics Melia 1 is therefore not about hidden fees. It is the difference between a direct vendor rate and a rate through a platform that also gives you one balance, one key and one usage log across video, audio and image.

If you only ever transcribe, and you do it in the hundreds of hours, a specialist's batch rate wins on cost. If transcription is one step in a pipeline that also generates video or speech, the extra amount is $0.48 per hour against Melia 1.

The job limit you need to plan for

Sume STT 1.0 takes at most 10 minutes of audio per request. A 500-hour archive is 30,000 minutes, so it needs at least 3,000 jobs if every file is cut to 10 minutes. Long recordings must be split before you submit them, and each piece is billed for its own minutes.

Splitting is cheap to script, and the Sume usage dashboard lets you check what you spent afterward. Speechmatics' page does not describe the same limit in the lines read for this post, so confirm file-length behaviour on their side before you assume parity.

A worked archive

Suppose the archive is 500 hours of 45-minute meeting recordings. That is about 667 files. On Sume each file must be cut into at most five pieces of 10 minutes or less (45 minutes is four pieces of 10 minutes and one of 5), so roughly 3,335 requests, billed at $0.01 per minute of audio actually submitted, which is still $300 in total because billing follows minutes and not request count.

On Speechmatics the file count does not change the per-hour rate, so the archive stays at $60, $115 or $190 depending on tier. The operational cost on the Sume side is the splitting step, and the saving on the Speechmatics side comes with an account, a key and a separate bill to reconcile.

Free credit changes the first month

Speechmatics' page offers $100 of free credits. Against Melia 1 that covers about 833 hours, so a first 500-hour archive could cost nothing on their side. Sume's Free plan is image generation only, so STT is billed from the first minute.

• Speechmatics Melia 1 with the $100 credit: $0 for 500 hours, if the credit applies to that tier.
• Speechmatics Standard: $115 before credit, so $15 after a $100 credit.
• Sume STT 1.0: $300, no free tier.

Rule of thumb

For a one-off archive in the hundreds of hours, price it on the vendor page and take the cheapest tier that meets your accuracy bar. For a recurring pipeline that already runs on Sume, the extra cost per hour is $0.48 over Melia 1, $0.37 over Standard and $0.22 over Enhanced. See Speechmatics' pricing for the live rates, and the Sume API pricing for ours.