28-minute podcast episode on Sume: narration, bed, render = $4.125

A 28-minute narrated episode costs about $4.125 on Sume: two TTS requests ($1.20), one Music 1.0 bed ($0.125) and a 28-minute Timeline render ($2.80).

5 min readSume
All posts

A 28-minute narrated episode costs about $4.125 on Sume when you publish it as a video with a cover still: $1.20 for the narration (two TTS requests), $0.125 for one Music 1.0 bed and $2.80 for the Timeline render. The only soft number is the script length, so the arithmetic below states the assumption and shows every step.

The assumption: how many characters is 28 minutes

Sume prices text-to-speech per character, not per minute, so you need a character count. This post assumes a planning figure of 900 characters per spoken minute. That is an assumption for budgeting, not a Sume figure; measure your own scripts. At 900 characters a minute, 28 minutes is 25,200 characters.

Narration: two requests because of the 20,000-character cap

Every Sonic model in the Sume TTS router accepts at most 20,000 characters per request. The list price is 38 micro-dollars per character and the billable price is the list price times 1.25, rounded up to the next cent for each job. That is 47.5 micro-dollars per character, and 10,000 micro-dollars make one cent.

Split the script at a paragraph break: 20,000 characters in the first request and 5,200 in the second. The first costs 20,000 x 47.5 = 950,000 micro-dollars, which is exactly 95 cents. The second costs 5,200 x 47.5 = 247,000 micro-dollars, or 24.7 cents, rounded up to 25 cents. Narration total: $1.20.

Narration cost for 25,200 characters (Sonic catalog, as of 2026-10-08)
RequestCharactersMicro-dollars at 47.5 eachBilled
120,000950,000$0.95
25,200247,000$0.25
Total25,2001,197,000$1.20

Music bed and render

Music 1.0 is a flat $0.125 per accepted generation. The price does not change with prompt length or an optional image. Music has no duration field, so you steer length in the prompt (for example, 'a 2-minute track'); a bed that is shorter than the episode can loop through the Timeline soundtrack settings.

Timeline 1.0 charges $0.10 per ceil(output minute). A 28-minute spine (1,680 seconds, under the 1,800-second cap) is 28 minutes, so $2.80. The two narration files can enter the render as audio.parts[], which accepts up to 20 gapless slices joined in the sample domain with no re-synthesis.

Full bill (as of 2026-10-08)
LineMathCost
Narration$0.95 + $0.25$1.20
Music 1.0 bed1 x $0.125$0.125
Timeline render28 x $0.10$2.80
Total$4.125

Where the money goes

The render is 68 percent of the bill: $2.80 of $4.125. That is the line to watch. Timeline rounds up to whole output minutes, so a 28-minute episode that runs 28 minutes and 5 seconds would be billed as 29 minutes ($2.90). Set audio.duration_seconds to the real length of the spine, and trim the tail silence before you plan the render. Narration is 29 percent ($1.20) and the bed is 3 percent.

If you publish audio only, you can skip the render and join the two narration files with Timeline audio concat, which returns one durable audio file. This post prices the video route because that is the one that needs the render line.

What this does not include

Cover art, captions, and any retakes are extra. A retake of a short line is usually one cent of TTS, since each job rounds up to a whole cent. If you only need audio and not a video file, skip the render line; the plan call can tell you the billable minutes before you spend anything.

Rerun it for your own show

Swap three numbers to price your own episode: characters (script length), the number of 20,000-character requests, and output minutes. For a 40-minute show the same assumption gives 36,000 characters, two requests ($0.95 + 16,000 x 47.5 = 760,000 micro-dollars, 76 cents) and a render that does not fit one Timeline call, since the cap is 1,800 seconds. Split such a show into two renders, as the related post on a 50-minute show explains.

Keep the arithmetic in a spreadsheet next to the episode. Prices come from the live catalog, so check GET /v1/catalog before you quote a client.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume