Google's September 22 release notes introduce Gemini 3.8 Flash TTS, Flash-Lite TTS, a voices endpoint, and custom-voice capabilities. The notes describe voice replication with consent verification. These are useful product features, but the organization still needs a policy for how voices are selected and used.
Match the voice to the job
Training narration, accessibility audio, marketing content, and live service responses have different requirements. Evaluate pronunciation, pacing, consistency over long passages, and behavior with technical terminology.
For interactive use, include generation latency and the handling of partial or corrected text. For produced content, assess editing effort and consistency across revised sections.
Treat a custom voice as a governed asset
Keep authorization and permitted-use records. Define whether the voice can be used after a contract ends, who may request new material, and how to revoke access. A supplier's verification step is not a replacement for your organization's rights review.
Avoid implying that a real person recorded a statement they did not record. Apply relevant disclosure requirements and brand standards.
Measure total production effort
Compare cost per accepted audio minute, including text preparation, quality checks, and regeneration. Test multilingual material with qualified reviewers rather than assuming a natural-sounding output is linguistically correct.
The right TTS service is not necessarily the one with the most expressive demonstration. It is the one that delivers understandable, authorized audio consistently within your production and governance requirements.
Sources & further reading
Reviewed October 26, 2026. Product availability and regulatory guidance can change. Vendor performance claims are not independent Rovio AI benchmarks.
Found this useful? Share it.