On September 23, 2026 local time, Google officially announced the launch of Gemini 3.8 Flash TTS and Gemini 3.8 Flash‑Lite TTS, two new text‑to‑speech models. The company positions these models as the most expressive audio‑generation models in the Gemini family to date, aiming to shift speech synthesis from traditional static preset voices to a deeply customizable, fine‑grained creative studio that meets developers' and creators' needs for personalized character dubbing and large‑scale audio generation.
Model positioning and core features
Gemini 3.8 Flash TTS is designed for deep creative applications and character design. Creators can use natural‑language prompts to build entirely new voices from scratch for game characters, immersive audiobooks, podcasts, and interactive media. The model supports line‑by‑line, fine‑grained control of acting cues, rhythm, dialect switching, and backchanneling, making voice generation closer to a real‑world directing workflow. It supports 130 languages and includes automatic language detection.
Gemini 3.8 Flash‑Lite TTS targets high‑throughput, cost‑sensitive, large‑scale scenarios. It is optimized for mass dubbing, audio content production, and expressive voice agents, offering precise control over tone, pacing, and subtle expressive differences. Language support reaches 101 languages, and it can also access more than 2,000 preset voices.
Both models support creating character voices via natural‑language descriptions (for example, "warm documentary narration" or "a quirky fantasy character with a regional accent") and allow fine‑tuning of timbre, perceived age, pitch, and texture. Flash TTS additionally supports voice cloning from approximately 30 seconds of audio samples, but requires recorded verbal consent from the voice owner; the system verifies that the sample matches the reference speaker before generating persistent or stateless voice configurations. All generated content is embedded with a SynthID invisible watermark, and cloned voices carry an additional C2PA content credential to ensure transparency and traceability.
These models join the Gemini Audio family, building on prior capabilities such as 3.5 Live Translate, 3.5 Transcribe, 3.8 Live, and 3.8 Live Extended Thinking to further strengthen the voice‑generation segment.
Performance and benchmarks
Google's published evaluations show that Gemini 3.8 Flash TTS ranked first overall (71.4 points) on Hume AI's Voice Design Benchmark and also led the accent modeling track (60.8 points). On the Overall Quality Index, Flash TTS and Flash‑Lite TTS ranked first and second, respectively. In human blind tests conducted by Voice Arena, both models placed near the top across key global languages including Japanese, Brazilian Portuguese, Vietnamese, Modern Standard Arabic, Mexican Spanish, and Hindi.
Compared with the Gemini 3.1 Flash TTS preview, the new models show significant improvements in long‑content stability, dual‑speaker script control, and nonverbal vocalizations (such as laughter and sighs), while maintaining high reliability.
Pricing and availability
Both models are available at promotional pricing through December 31, 2026; standard pricing applies from January 1, 2027. Promotional rates are: Flash TTS text input $0.50 per million tokens and audio output $9.00 per million tokens; Flash‑Lite TTS input $0.50 per million tokens and output $6.00 per million tokens. Batch and Flex modes are half price, while Priority mode is about 1.8×. The new pricing is notably more developer‑friendly compared with the 3.1 Flash TTS preview rates of $1.00/$20.00.
Developers can access the models immediately via the Gemini API and Google AI Studio; Flash TTS is also available in Gemini Notebook, and Flash‑Lite TTS is integrated into Google Vids. An enterprise‑level API will be available soon in Gemini Enterprise.
Potential impact
This release substantially lowers the barrier to creating high‑quality, personalized voice content. Game developers, audiobook producers, podcasters, and interactive media creators can design character voices and performance scenes more freely without relying on limited preset libraries or costly bespoke recordings. Large‑scale dubbing and voice‑agent scenarios benefit from Flash‑Lite's efficiency and cost advantages, supporting scalable multilingual content production.
For enterprises, the combination of fine‑grained performance control and reliable watermarking improves creative flexibility while preserving provenance, aiding compliant use of generated audio. For the broader speech‑AI field, Google's benchmark‑leading performance and open natural‑language voice design further compress the space for traditional TTS and competitors, potentially accelerating the industry's shift toward voice systems that are directable, clonable, and scalable.
At the same time, post‑promotion pricing adjustments, enterprise adoption timelines, and real‑world latency and stability in production environments will determine long‑term adoption. Overall, the Gemini 3.8 Flash TTS series advances text‑to‑speech from a "reading" tool to a "creative performance studio," providing developers and creators with a more powerful and flexible speech‑generation infrastructure.
▶️ Enjoyed this article? Creating takes real effort!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions, and shares are our greatest support! 🙏
For more AI news visit: https://www.airosetta.com