Gemini 2.5 Flash Preview TTS: Two-Speaker Audio Generation
Google released Gemini 2.5 Flash Preview TTS on May 20, 2025. This model generates audio from text for one or two speakers with an 8192-token context window.
Official Release Details
Google officially released the gemini-2.5-flash-preview-tts model on May 20, 2025. This announcement occurred simultaneously with the release of the gemini-2.5-pro-preview-tts model. Both models are categorized as previews within the Gemini family. The primary function documented in the official changelog is the generation of speech capable of utilizing one or two speakers.
- Release date: May 20, 2025
- Model identifier: gemini-2.5-flash-preview-tts
- Input modality: Text
- Output modality: Audio
Technical Specifications and Limits
The verified technical specifications for the Gemini 2.5 Flash Preview TTS model indicate a context window of 8192 tokens. The maximum output length is set at 16384 tokens. The source data records that open weights are not confirmed for this model. No pricing information is provided in the available facts. The documentation does not specify deployment architecture, local installation requirements, or specific integration behaviors beyond the API observation.
- Context window: 8192 tokens
- Max output tokens: 16384 tokens
- Open weights status: Not confirmed
- Pricing: Not specified
Evaluation Procedure for Engineers
Engineers can evaluate the model by designing tests that validate its core capability of generating speech for one or two speakers. A practical evaluation procedure involves creating distinct text prompts designed to trigger single-speaker and dual-speaker outputs. Measure the latency and audio quality of the generated files. Compare the consistency of speaker differentiation between the two modes. Validate that the output remains within the 16384-token limit for long-form inputs. Since the source does not confirm performance rankings or guaranteed operational benefits, evaluations should focus solely on verifying the stated functionality against the provided test cases.
Primary sources
Frequently asked questions
Does the Gemini 2.5 Flash Preview TTS support multiple speakers?
Yes, the official release notes state that the model is capable of generating speech with one or two speakers.
Are open weights available for this model?
No, the source facts record that open weights are not confirmed for this release.