Gemini TTS API polling: the Sume audio job pattern

The Gemini changelog for September 22, 2026 announces gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts as generally available, but the entry does not describe polling. For polling you need a job API. On Sume, text-to-speech is a job: submit with mode: "async", keep the job id, poll GET /v1/jobs/:id/status until terminal is true, then read the result. If a wait or your process ends early, poll again; do not resubmit.

Gemini facts are from its changelog; the job pattern is from Sume's Jobs and results, read 2026-10-01.

What did Gemini announce for TTS?

Two GA models: gemini-3.8-flash-tts, described as the flagship creative TTS model, and gemini-3.8-flash-lite-tts, described as cost-efficient and meant to replace gemini-3.1-flash-tts-preview. The same entry adds a Voices endpoint at /v1beta/voices. The entry is a release note, so how a request completes is not covered there.

How does polling work on Sume?

A submit is accepted once Sume has a durable job id, and every mode returns that id in the first response. A 2xx means the job exists and paid work is in flight, not that it finished. The docs tell you to store the id "so your integration can recover work after process restarts", and to use exponential backoff, stopping on completed, failed or canceled.

const headers = { Authorization: "Bearer " + process.env.SUME_API_KEY };

async function waitForJob(jobId) {
  let delayMs = 2000;
  for (;;) {
    const res = await fetch("https://api.sume.com/v1/jobs/" + jobId + "/status", { headers });
    const s = await res.json();
    if (s.terminal) return s;
    delayMs = Math.min(delayMs * 2, 15000);
    await new Promise((r) => setTimeout(r, (s.next_poll_after_seconds ?? delayMs / 1000) * 1000));
  }
}

Why not resubmit when a call times out?

Because the first submit already created a paid job. The docs say not to resubmit the original paid request just because a local process timed out. If the submit call itself failed and you do not know whether it landed, retry with the same Idempotency-Key so you get the original job back rather than a second one.

Where do voices and engines come in?

Voice and engine choice happen at submit time and do not change the polling loop. See finding a voice id on Sume for voices and which engine Sume routes for Gemini TTS for engine selection.