Google Releases Gemini 2.5 Pro Preview TTS
Google announced the Gemini 2.5 Pro Preview TTS model on May 20, 2025. This text-to-speech preview supports one or two speakers and has an 8192-token context…
Official Announcement Details
Google officially announced the release of the gemini-2.5-pro-preview-tts model on May 20, 2025. This announcement occurred alongside the release of the gemini-2.5-flash-preview-tts model. Both models are identified in the official changelog as capable of generating speech with one or two speakers. The event is recorded as a preview release within the Gemini family.
Documented Technical Specifications
The source facts establish specific technical parameters for the Gemini 2.5 Pro Preview TTS model. The input modality is strictly text, while the output modality is audio. The model operates with a context window of 8192 tokens and allows for a maximum output of 16384 tokens. The provider is listed as Google. Regarding licensing, openWeights is marked as false, which records that open weights are not confirmed for this specific model iteration. No pricing information is provided in the source data.
- Input Modality: Text
- Output Modality: Audio
- Context Window: 8192 tokens
- Max Output Tokens: 16384 tokens
- Open Weights: Not confirmed
Evaluation Procedure for Engineers
Engineers can evaluate the model by designing controlled tests to validate its speech generation capabilities. First, prepare a set of standardized text prompts that require single-speaker and dual-speaker outputs. Second, measure the latency and audio quality of the generated responses against existing baselines. Third, compare the consistency of speaker differentiation between the two supported modes. Finally, validate that the output remains within the documented 16384-token limit during extended generation tasks. These steps allow for a practical assessment of the model's performance without assuming unverified architectural benefits.
Primary sources
Frequently asked questions
What date was the Gemini 2.5 Pro Preview TTS model announced?
The model was announced on May 20, 2025. This historical date is distinct from the current publication date of this article.
Does the source confirm if local installation is possible?
No. The source states that open weights are not confirmed (openWeights=false). It does not establish whether local installation is impossible, nor does it confirm any specific hosting arrangement or API availability details beyond the model's existence in the API.