Six podcast ad-read variants at 500 characters: TTS cost and tags

Six podcast ad-read variants of 500 characters each cost $0.1425 in total on Sume TTS, which is $0.02375 per read at $0.0475 per 1,000 characters. That makes it cheap to voice every headline you want to test and let the sponsor pick.

Cost of the set

The per-read price is the price tool output for 500 characters. Totals are that number times the count.

Tag every take so you can find the winner

TTS accepts a metadata object. Sume stores it on the job and does not send it to the provider, so it is a safe place for your variant label, sponsor and campaign id. Use one Idempotency-Key per variant (for example ad-q4-A, ad-q4-B) so a retry of variant C cannot create a second C.

Reuse speak() from the previous post pattern, but pass your variant id into the key and add metadata to the body. Only the body differs:

body = {
    "transcript": variant_text,          # 500 characters
    "avatar_handle": "@narrator",
    "language": "en",
    "metadata": {"campaign": "q4-sponsor", "variant": "B"},
}
requests.post("https://api.sume.com/v1/tts-1.0/generate", json=body,
              headers={**H, "Idempotency-Key": "ad-q4-B"}).raise_for_status()

What to check before you send six

Keep the same voice, language and generation_config across variants, otherwise you are testing the voice and not the copy. generation_config takes volume (0.5 to 2), speed (0.6 to 1.5) and a free-text emotion up to 64 characters.

Read the sponsor's required legal line into every variant at the end; it counts toward the 500 characters. If a read needs to be exactly 30 seconds, generate once, measure the audio duration, then adjust speed on the retake rather than cutting words.