Google Gemini 3.5 Transcribe: the real cost and what they don't tell you

Unveiling the Power and Fine Print of Gemini 3.5 Transcribe

On August 26, 2026, Google launched Gemini 3.5 Transcribe, a state-of-the-art model designed for audio transcription. This release included two key variants: one optimized for pre-recorded files and another for real-time streaming. While Google advertises an attractive price of $0.30 per hour of audio, there are crucial nuances that every potential user should be aware of to avoid surprises and optimize costs.

Advanced Capabilities and Realistic Pricing

Gemini 3.5 Transcribe offers a robust set of functionalities that distinguish it in the market:

• Per-utterance language detection: Support for over 85 languages, identifying language switches within the same conversation.
• Speaker diarization: Allows identifying who said what, ideal for meetings and interviews.
• Word-by-word timestamps: Precise temporal location of each term in the audio.
• Custom vocabulary: Up to one thousand specific terms to improve accuracy in specialized domains.

The listed price of $0.30 per hour may seem straightforward, but the operational reality is more complex. Google's own documentation reveals that activating certain advanced features, such as diarization or word-by-word timestamps, can significantly reduce processing limits. For example, a one-hour limit can drop to thirty minutes, effectively doubling the hourly cost for those functionalities.

Optimizing Transcription for Business Use Cases

Understanding the models and their limitations is crucial for business applications. Confusing the file model with the streaming model can lead to unexpected costs, as the former has a one-hour limit and the latter a ten-minute per-session limit. Here we explore how to apply these technologies in common scenarios:

• Meeting minutes: Diarization transforms transcription into structured minutes, where the real value lies in speaker attribution, not just the text.
• Automated subtitles: Word-by-word timestamps eliminate the need for manual subtitle alignment, saving time and improving accessibility.
• Call analysis: Integrating custom vocabulary allows transcribing and analyzing support or sales calls with your company's specific terminology, improving insight extraction and AI training.

How to apply it in your business?

To leverage Gemini 3.5 Transcribe, it's essential to understand its two models and how they interact with your specific needs. Evaluate whether your use case requires diarization or timestamps and adjust your cost strategy accordingly. Consider if Google's precision and advanced features justify the incremental cost versus alternatives like Whisper, which, while free, lacks certain functionalities and support. The key is to design the solution so that the output (like meeting minutes or subtitles) is the valuable final product, not just raw transcription.