A birthday voice message with a music sting for about 14 cents on Sume

The recipe and the total

Generate the greeting with Sume TTS, generate a short music piece with the music router, and join them in order with timeline audio concat. A 200-character greeting is 0.2 x $0.0475 = $0.0095, one music generation is $0.125, and the join is $0.01 flat, so the whole message is $0.0095 + $0.125 + $0.01 = $0.1445.

The join is a concatenation, one file after the other. It is not a mix, so the music plays first and the voice follows, or the reverse. Design the message around that: a bright sting, then the greeting.

Cost by step

Every number is a Sume list price from the docs or the golden pricing file in the repository.

The three requests

First, the greeting. Write it as it should be heard, with the name spelled as spoken: "Happy birthday, Maya. Hope the day is full of cake and good news." Submit it with a warm voice and a mild emotion string, and language set.

Second, the sting. Ask for a short piece in the prompt, for example "a bright 10-second birthday sting, acoustic guitar and handclaps, 110 BPM, G major, ends with a clear final chord. Instrumental, no vocals." The music router docs say length goes in the prompt, because a duration field is rejected.

Third, join them with POST /v1/timeline-1.0/audio using operation: "concat" and the two audio URLs in order. The timeline audio docs explain that the join is gapless and needs no re-synthesis.

Before you send it

Listen to the finished file. If the sting is longer than you wanted, regenerate it with a tighter prompt; the greeting does not need to change, and you only pay for the part you redo. Keep the three job ids with the final file.

• Write names the way they should sound.
• Keep the greeting under 20 seconds so it holds attention.
• Ask the recipient's permission before you use a real person's recorded voice; this recipe uses a stock voice.
• Do not name a living artist in the music prompt.

Variations

Put the voice first and the music last for a sign-off, or join music, voice and music for a bookend. Each extra join is within the same single concat job, which accepts 1 to 20 parts at $0.01 flat, so the cost stays $0.01 however many pieces you stack. A 20-second greeting at about 400 characters costs $0.019, which makes the voice the cheapest part of the whole message.

Cost of trying it three ways

If you try three sting prompts and keep one, the music costs 3 x $0.125 = $0.375, and the message total rises to $0.3945 with the voice and the join. That is still under forty cents for a personal message, and the variety is worth more than the saving from stopping at the first take.