Phrase-of-the-day language video: a card and a voice for $0.21

A phrase-of-the-day video is one card image, one spoken line and one Timeline render: $0.10 for a Nano Banana 2.1 card at 1K, $0.0057 for a 120-character TTS line and $0.10 for a render of up to a minute, so $0.2057 per phrase and $1.44 for a week of seven.

Cost for a week

Add $0.20 per video if you also want burned-in captions. Both sums use the exact arithmetic below.

The voice call

Set language for every non-English line; otherwise the voice defaults to English. Ask for wav, and timestamps.words if you want word times for the card. For Korean captions, choose a Hangul style.

curl -X POST https://api.sume.com/v1/tts-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: phrase-2026-10-09-ko" \
  -d '{
    "transcript": "Gamsahamnida. Thank you very much.",
    "avatar_handle": "teacher_host",
    "language": "en",
    "output_format": { "container": "wav", "sample_rate": 44100 },
    "timestamps": { "words": true }
  }'

Limits and gotchas

• One TTS job tops out at 20,000 characters, and audio over 1,200 seconds fails.
• A voice speaks one language per request; for mixed lines use one call per language and join the audio with Timeline audio concat ($0.01).
• Check pronunciation yourself. A voice can mispronounce a rare word.