Google Gemini 2.5 Pro preview with TTS, accessed via OrcaRouter at zero markup.
google/gemini-2.5-pro-preview-tts is a preview text-to-speech model by Google, built on the Gemini 2.5 Pro foundation. It converts text input into spoken audio output. The model is accessed through…
The primary capability of google/gemini-2.5-pro-preview-tts is converting written text into spoken audio. It is built on Google's Gemini 2.5 Pro, indicating strong language understanding, which should result in natural phrasing, appropriate intonation, and clear pronunciation. The model accepts text input only, so it is not multimodal on the input side. It does not handle audio input, images, or video. The output is likely audio in a standard digital format. As a TTS model, it can handle multiple languages, but specific language support has not been disclosed for this preview. It is optimized for generating speech, not for other Gemini 2.5 Pro capabilities like reasoning or code generation.
The model excels in applications requiring conversion of written text to natural-sounding speech. Best use cases include voice assistants where the assistant reads responses aloud, automated audiobook and podcast narration, accessibility features that read on-screen text for visually impaired users, and customer service systems that deliver spoken responses. It can also be used for language learning applications where correct pronunciation is modeled, or for generating voice content from RSS feeds or articles. Because of its preview status, it is advisable to test thoroughly for your specific language, domain, or style requirements before committing to large-scale deployment.
While google/gemini-2.5-pro-preview-tts offers premium TTS quality, it may be overkill for simple beeps or notification sounds, or for applications where low latency is critical but the model's processing time is longer than dedicated TTS chips. If you need only basic, monotone speech or have very high volume with strict budget constraints, a lighter TTS model (e.g., from other providers on OrcaRouter) with lower per-token costs may be more appropriate. Also, if your use case requires real-time streaming with extremely low latency, evaluate whether Gemini's preview model meets your speed requirements, as larger models can introduce higher first-token latency.
As of the model's preview release, Google has not published specific benchmark scores for the gemini-2.5-pro-preview-tts variant. The underlying Gemini 2.5 Pro base model has demonstrated strong performance on language understanding and reasoning tasks, but the TTS variant's speech quality metrics (e.g., Mean Opinion Score, Word Error Rate) have not been publicly shared. Without official benchmarks, OrcaRouter recommends evaluating the model on your own test set of text inputs, measuring subjective audio quality, latency, and reliability under load. For production decisions, conduct your own A/B comparisons against other TTS services.
Specific latency figures for google/gemini-2.5-pro-preview-tts have not been publicly disclosed. As a TTS model based on a large language model architecture, it may exhibit higher latency compared to specialized neural TTS models that are optimized for real-time streaming. Factors that influence speed include the length of the input text, the model's internal processing time, and the network round-trip to OrcaRouter's endpoints. For applications where low latency is critical (e.g., conversational AI), test this model with typical input lengths to gauge its suitability. Consider using streaming or chunked text generation if the API supports it.
Strengths: Built on Google's advanced Gemini 2.5 Pro architecture, it likely produces highly natural, expressive speech with good prosody and pronunciation. Access via OrcaRouter's OpenAI-compatible API simplifies integration. No markup on provider pricing makes it cost-predictable. Limitations: Input modality is text-only; it cannot process audio or other modalities. Being a preview, it may have undiscovered edge cases or inconsistent quality. The context window size is unknown, restricting the length of text that can be converted in a single request. Output format details are not provided. It is not designed for tasks like speech recognition or voice cloning.
Pricing for google/gemini-2.5-pro-preview-tts is based on token usage, charged at $1.00 per 1 million input tokens and $20.00 per 1 million output tokens. Input tokens include the text prompt and any instructions. Output tokens represent the generated audio. OrcaRouter bills at the exact provider rate with zero markup, meaning you pay only what Google charges, plus any applicable taxes. There are no minimum commitments or monthly fees. Usage is metered per request and aggregated on your OrcaRouter invoice. Because TTS output token counts can be high (audio being more token-dense than text), output cost is the primary expense.
The output token price ($20 per 1M) is significantly higher than input tokens ($1 per 1M). This is typical for generative models because output tokens require more compute. For text-to-speech, the audio output is represented as a series of tokens, and converting text to speech involves generating a long sequence of audio tokens. The total cost for a specific task depends on the length of the output audio relative to the input text. If you need to generate long stretches of speech (e.g., a full audiobook), costs can accumulate. For short prompts (e.g., a single sentence), costs are minimal. Monitor your token usage to project costs.
No, OrcaRouter does not add any markup to the provider rate for google/gemini-2.5-pro-preview-tts. The listed prices of $1.00 per 1M input tokens and $20.00 per 1M output tokens are exactly what Google charges. OrcaRouter passes these through to you directly. This is consistent with OrcaRouter's pricing model for many models where the platform acts as a transparent proxy. You only pay for the tokens you consume. Any optional features such as caching or premium support may incur separate fees, but the base per-token cost has no markup.
To use google/gemini-2.5-pro-preview-tts, make requests to OrcaRouter's OpenAI-compatible API at base URL https://api.orcarouter.ai/v1. Use the model ID 'google/gemini-2.5-pro-preview-tts' in your API calls. The request format follows the standard OpenAI chat completions structure with a 'model' field, a 'messages' array containing your text prompt (with a system and/or user role), and standard parameters like temperature and max_tokens. The response will contain the generated audio tokens, which your client can decode to play as speech. Authentication is done via an API key provided by OrcaRouter.
The API supports standard OpenAI-compatible parameters including 'model', 'messages' (array of objects with role and content), 'temperature' (for controlling audio output variability), 'max_tokens' (to limit the length of generated audio), and 'top_p'. As a TTS model, there may be additional parameters for voice style or speed, but these have not been documented in the available facts. Refer to OrcaRouter's API documentation for the latest supported fields. The 'response_format' parameter may allow specifying audio encoding (e.g., wav or mp3). If not supported by default, you may need to decode the returned token stream.
If you are currently using a different TTS service (e.g., Google Cloud Text-to-Speech, Amazon Polly), migrating to google/gemini-2.5-pro-preview-tts via OrcaRouter involves updating your API endpoint and request format. OrcaRouter uses OpenAI-compatible endpoints, so you'll switch from a provider-specific SDK to an OpenAI-compatible one. Replace your existing base URL with https://api.orcarouter.ai/v1, use the model ID 'google/gemini-2.5-pro-preview-tts', and adjust your request bodies to match the chat completions schema. You may need to modify how you handle the response: instead of receiving audio files directly, you'll get tokenized output that you'll need to decode. Test with sample inputs.
Authentication is done by including an API key in the Authorization header (format: Bearer YOUR_API_KEY). You obtain the API key from your OrcaRouter account. Rate limits are set by OrcaRouter based on your plan; they are not the same as Google's limits. For high-volume usage, contact OrcaRouter to adjust limits. The preview model may have additional constraints from Google – if you encounter errors such as 'model not available', check that your OrcaRouter plan includes this model. There are no specific rate limits disclosed for this model, but standard best practices (retry with exponential backoff) are recommended.
google/gemini-2.5-pro-preview-tts is a specialized variant of the Gemini 2.5 Pro family tailored for text-to-speech. Other Gemini 2.5 Pro models are general-purpose and handle text, images, audio, and video input; they do not natively generate speech output. This TTS variant strips out multimodal input and focuses on generating audio from text. It likely uses the same underlying language understanding for prosody and pacing, but it is not suitable for reasoning, coding, or multimodal analysis. If you need TTS capabilities without giving up multimodal input, you would combine a standard Gemini model with a separate TTS pipeline. The preview TTS provides a more streamlined pipeline.
OrcaRouter offers access to TTS models from various providers. This model from Google is based on a large language model architecture, which may produce more natural and contextually aware speech than older concatenative or parametric TTS systems. However, it may be slower and more expensive per token compared to specialized neural TTS models designed for low latency and high throughput. Many popular TTS services charge per character or per second of audio, not per token, making direct cost comparison tricky. For production deployments requiring extremely low cost, dedicated TTS models may be preferable. Google's model is best for use cases where the highest output quality justifies the higher output token cost.
Choose this model if you need state-of-the-art speech quality from Google’s latest Gemini architecture and you want to access it via a standard OpenAI-compatible API without managing Google Cloud credentials. The zero-markup pricing ensures transparency. It is ideal for applications where naturalness is critical, such as conversational AI or narrative content. The preview status allows early experimentation to evaluate its quality for your domain. However, if you need real-time streaming with sub-100ms latency, a dedicated streaming TTS model might be better. Also, if your budget is sensitive, test the output token consumption for typical use cases to ensure the cost is acceptable.
OpenAI-compatible — keep the SDK you already use
https://api.orcarouter.ai/v1voice| Input / 1M tokens | $1.00 |
| Output / 1M tokens | $20.00 |
| Currency | USD |
Estimate based on list price
Estimate only — actual token counts depend on the provider's tokenizer.
What developers are saying this week
GET /api/public/models/google/gemini-2.5-pro-preview-ttsOpen @misc{orcarouter_gemini_2_5_pro_preview_tts,
title = {google/gemini-2.5-pro-preview-tts API},
author = {google},
year = {n.d.},
howpublished = {OrcaRouter},
url = {https://www.orcarouter.ai/models/google/gemini-2.5-pro-preview-tts}
}google. (n.d.). google/gemini-2.5-pro-preview-tts API. OrcaRouter. https://www.orcarouter.ai/models/google/gemini-2.5-pro-preview-tts