Voice replication (POST /v1beta/voices) always returns 404 "Requested entity was not found" — normal TTS works

Gemini API
1

Hi, I’m trying voice replication with gemini-3.8-flash-tts following
https://ai.google.dev/gemini-api/docs/voice-replication, but creating a voice always fails with:

404 NOT_FOUND “Requested entity was not found.”

Normal TTS on the same key works fine (POST /v1beta/interactions with gemini-3.8-flash-tts returns audio),
and GET /v1beta/voices returns the prebuilt voice list. Only voice creation fails.

Environment

  • Region: Japan
  • Paid Tier 1, prepaid credits available (balance confirmed in AI Studio billing)
  • Tested on two projects: an existing project and a brand-new project created in AI Studio today
    with billing set up and a new API key. Same result on both.

What I tried (all return the same 404)

  • REST: POST https://generativelanguage.googleapis.com/v1beta/voices with
    {“store”: true, “voice”: {“model”: “gemini-3.8-flash-tts”, “type”: “replicated”,
    “replicated”: {“source_audio”: {…}, “consent_audio”: {…}}}}
  • Official Python SDK google-genai (>= 2.25.0): client.voices.create(…)
  • AI Studio UI: “Create new voice” > Voice Replication > “Replicate voice” shows the same
    “Requested entity was not found.” toast
  • store: true / false, with and without display_name, omitting voice.model
  • Audio: m4a (audio/mp4) and 24kHz mono 16-bit WAV (audio/wav)
  • Source audio 19-29 sec; consent audio 13-16 sec; recorded with the same mic in the same room
  • Consent phrase read verbatim in ja-JP and in en-US (both tried)
  • Re-recorded both clips from scratch; same result

Possible doc mismatch

  • The docs say gemini-3.8-flash-lite-tts supports voice replication, but the API returns
    400 “Model ‘gemini-3.8-flash-lite-tts’ does not support Voice Replication”.

Question
Is voice replication fully rolled out to all paid accounts/regions yet, or is there an additional
requirement (allowlist, account setting, region) not mentioned in the docs?
Since the same 404 occurs in the AI Studio UI, it seems to be account- or region-side rather than
a request format issue.

Thanks!

2

Update: this is resolved on my side.

Voice creation started working again on October 2, 2026 (around 03:00 UTC), with the same recordings, project and request as before — no changes on my end. The voice was created successfully and TTS with the replicated voice works.

So this was the same backend issue reported in the other thread (404 on CreateVoice since Oct 1), not a region or account restriction. Thanks to the team for the fix.