Furnace tune-up reminder video: 3 stills, voiceover, 61 cents

A heating company can make a furnace tune-up reminder video with no video model at all: generate three stills, record a 292-character voiceover with Sume TTS (about 1.4 cents), lay the stills over the voice with one Timeline 1.0 render ($0.10), and burn captions from the speech ($0.20). With three Nano Banana 2.1 stills at about $0.10 each, the whole video costs about $0.61. It suits a trade that wants a clear, calm, 15-second reminder in the weeks before the first hard freeze.

Why stills and not generated video?

A reminder has one job: get a booking. Three still scenes (the furnace closet, a technician's gloved hands with a filter, a warm living room) and a clear voice carry that. Stills in Timeline 1.0 are static holds, per the Timeline docs (read 2026-10-11): motion on a still is accepted and ignored, with a motion_ignored warning, so do not count on a push-in. If you want movement, generate a clip instead.

What is the script, and what does it cost?

Sume TTS is billed per transcript character at $0.0475 per 1,000, spaces and punctuation included. This script is 292 characters, so the voiceover is about $0.0139.

The first hard freeze is coming. If your furnace has not been checked this year, now is the time. A tune-up takes about an hour. We clean the burners, test the safety controls, and change the filter. Book before the cold snap, when our schedule fills up. Call, or tap the link to pick a time.

Read it aloud before you generate. Roughly 15 seconds is right; if the audio comes out at 17 seconds, your timeline length must be 17, not 15.

How do I lay the stills over the voice?

The timeline needs audio.duration_seconds and a Sume-hosted audio file, and every slot must be on media.sume.com. Run the free plan call first (POST /v1/timeline-1.0/plan); it returns the length, the segment count and the estimated cost without creating a job. Slot starts must increase from 0, and coverage must reach within half a second of the end of the audio. Transitions apply only after the first slot.

A 15-second voice and three 5-second slots look like this:

{
  "audio": { "url": "https://media.sume.com/artifacts/artf_demo/voice.mp3", "duration_seconds": 15 },
  "video": [
    { "source_url": "https://media.sume.com/artifacts/artf_demo/closet.png", "start": 0, "duration": 5 },
    { "source_url": "https://media.sume.com/artifacts/artf_demo/filter.png", "start": 5, "duration": 5,
      "transition": { "type": "fade", "duration": 0.5 } },
    { "source_url": "https://media.sume.com/artifacts/artf_demo/living-room.png", "start": 10, "duration": 5,
      "transition": { "type": "fade", "duration": 0.5 } }
  ],
  "output": { "fade_out_seconds": 0.5 }
}

Where do captions go, and what is the total?

Run the caption job on the rendered MP4. Because the voice is real speech, you do not need cues; STT does the timing, and script_text can align the burned text to your script, which helps with a company name. The standalone job is $0.20 for a video up to 60 seconds. Timeline 1.0 is $0.10 per output minute, rounded up, so 15 seconds bills as one minute.