Google released Gemini 3.1 Flash TTS on April 15, 2026, a text-to-speech model that outperforms ElevenLabs v3 in quality benchmarks while offering a generous free tier. The model brings expressive, multilingual voice generation to creators who have been waiting for a serious Google entry in the AI voice space.

What Happened

Google launched Gemini 3.1 Flash TTS through the Gemini API and Google AI Studio. Built on the Gemini 3.1 Flash architecture, it supports over 70 languages and introduces audio tags, a simple text-based control system that lets developers direct style, tempo, tone, and accent within a single prompt. All generated audio ships with Google SynthID watermarking to identify AI-generated content.

According to Artificial Analysis rankings, Gemini 3.1 Flash TTS earned an Elo rating of 1,211, placing it above ElevenLabs v3 in overall quality, just behind Inworld 1.5 Max. On pricing, the paid tier costs $1.00 per million input tokens and $20.00 per million audio output tokens, with 50% off in batch mode. A free tier is available, with the tradeoff being that Google may use that data for product improvement.

Why It Matters

For creators building voiceover workflows, narration pipelines, or multilingual content, this changes the competitive landscape. ElevenLabs has been the benchmark for expressive AI voice for the past two years. A Google model that beats it on quality, at $1 input and $20 audio output per million tokens, is a meaningful shift. The batch mode discount brings that down to $0.50 and $10.00 respectively, which is competitive for high-volume production work.

The audio tag system is worth paying attention to. Instead of engineering complex prompt workarounds to get the right tone, creators can now specify style and tempo inline. That kind of control has historically required fine-tuned models or multiple API calls to iterative services like ElevenLabs.

Key Details

  • Elo rating: 1,211 on Artificial Analysis TTS leaderboard
  • Quality ranking: Above ElevenLabs v3, behind Inworld 1.5 Max
  • Languages: 70+ supported
  • Multi-speaker: Yes, handles multi-speaker dialogs natively
  • Watermarking: SynthID applied to all outputs
  • Free tier: Available via Google AI Studio with data use tradeoff
  • Paid pricing: $1.00/M input tokens, $20.00/M audio output tokens
  • Batch mode: 50% discount ($0.50 and $10.00 respectively)
  • Available via: Google AI Studio, Gemini API, Vertex AI, Google Vids

What to Do Next

The free tier at aistudio.google.com/generate-speech is the fastest way to test it. Run a few samples in your target language and compare directly against your current voice pipeline. Pay attention to the audio tag controls, setting a calm, deliberate pace versus energetic delivery is now a matter of a few words in the prompt rather than swapping models. For high-volume projects, calculate whether the batch tier is cheaper than your current provider before committing to a switch.

Creators building multilingual content pipelines have a clear reason to test this immediately. The 70+ language support, combined with benchmark-leading quality, makes this a strong candidate to replace or supplement existing voice tools. If you are already using open-source TTS alternatives for cost control, the free tier here may be all you need for prototype work.