45-second teaser from a 12-minute podcast on Sume: $0.43
STT to find the quote, split 45 seconds, render over a cover still, burn captions: $0.12 + $0.01 + $0.10 + $0.20 = $0.43. Each step and its limit.

A 45-second teaser from a 12-minute podcast on Sume is four steps: transcribe to find the quote ($0.12), split a 45-second range ($0.01), render it over one cover still ($0.10), and burn captions ($0.20). The total is $0.43. If you already know the timestamps, skip the transcript and the cost is $0.31.
The four steps
Sume STT is $0.01 per audio minute, and it reserves for at most 10 minutes per job, so a 12-minute episode is two jobs of 6 minutes: 12 x $0.01 = $0.12. Timeline audio split cuts a range and returns a new media.sume.com file for $0.01. The render takes that file as audio.url with a 45-second audio.duration_seconds and one video[] slot holding the cover still, and reserves ceil(45 / 60) = 1 minute at $0.10. A standalone caption job is $0.20 for a clip up to 60 seconds, and runs speech-to-text on the clip itself.
| Step | Surface | Basis | Cost |
|---|---|---|---|
| 1 Find the quote | STT 1.0, 2 jobs of 6 minutes | 12 audio minutes x $0.01 | $0.12 |
| 2 Cut the audio | Timeline audio split, range 190-235 s | flat per job | $0.01 |
| 3 Make the video | Timeline render, still + audio | 1 minute x $0.10 | $0.10 |
| 4 Burn the captions | Video captions, 45 s clip | fixed per job | $0.20 |
| Total | $0.43 |
Details at each step
Four details keep the chain from breaking.
- Import the episode first. Split reads only this workspace's media.sume.com audio and rejects off-host URLs.
- Pick the range from the STT
words[]times, with a 0.2-second lead-in, so the clip starts on a breath, not a syllable. - In the render, a still is a static hold. Set the slot's
startto 0 and itsdurationto 45; coverage may stop at most 0.5 seconds before the spine ends. - Caption the rendered MP4 last. The caption job needs speech in the clip; a clip with no speech fails with
caption_no_speech.
Choosing the style and checking the cost
Captions default to a style chosen from the text: slam for Latin text, black-outline for Korean. For a quote card you may prefer punch or tiktok-green; those two do not take design overrides. Because source_caption_id lets you restyle the same clip without transcribing again, a second look is a second $0.20 render, not a new transcription.
Run the unbilled timeline plan before the render to confirm billable_minutes. A teaser that is 61 seconds long would reserve 2 minutes instead of 1, doubling that step to $0.20.
When to skip a step
If your editor already gives you the in and out times, skip step 1 and pay $0.31. If the audio is already a clip, skip step 2. If the platform burns its own captions, skip step 4. Each skipped step is a job you did not need, and the arithmetic above is the whole bill: there is no per-seat or per-episode fee on these surfaces beyond the rates in the table.
Sources
Related posts
More in Use cases
- 54-card deck art at 2:3: $2.36 on Seedream 5 Lite, $5.40 on Banana
Art for a 54-card deck at 2:3 costs $1.35 to $5.40 on Sume depending on the model. GPT Image 2.5 has no named 2:3 ratio; a custom 1024x1536 size is legal.
- 60-second explainer Short end to end: $7.84 with Wan 3.0 clips
A 60-second explainer built from TTS, twelve 5-second Wan 3.0 clips, one Timeline render and captions costs $7.84 on Sume. The clips are $7.50 of it.
- 63-slide deck, one 16:9 image per slide: $1.56 to $9.45 on Sume
A 63-slide deck with one 16:9 image each costs $1.56 on GPT Image 2.5 low, $2.76 on Seedream 5 Lite, $6.30 on Nano Banana 2.1 at 1K and $9.45 at 2K.
- 7-minute training video: free Timeline plan, then a $0.70 render
Preflight a 7-minute training video on Sume: POST /v1/timeline-1.0/plan is unbilled and returns billable_minutes. The render then costs 7 x $0.10 = $0.70.
Written by Sume