Gemini STT

**Verdict:** A focused, practical Gemini/Vertex AI speech-to-text CLI skill that is well-suited for Clawdbot integrations and scripting, with solid format support and simple auth handling. It is strongest when you already use Google Cloud or Gemini and need fast, lightweight transcription. **Best use cases:** - Converting Telegram or other voice messages to text inside bots or automation pipelines. - Quickly wiring Gemini/Vertex AI transcription into shell scripts or backend services. - Environments where installing extra Python packages is undesirable and ADC (gcloud) is available. **Key strengths:** - No external Python dependencies; easy to drop into most Python 3.10+ environments. - Flexible authentication (ADC-first, then API key) with clear CLI options for forcing Vertex, project, and region. - Sensible default model (`gemini-2.0-flash-lite`) optimized for speed and cost, while still allowing higher-quality models. **Notable limitations / risks:** - Depends entirely on Google’s Gemini/Vertex AI APIs: subject to model changes, quota limits, and network latency; offline or on-prem use is not supported. - The description does not mention advanced transcription features (timestamps, diarization, confidence scores, very long audio chunking), so it appears to provide plain-text output only. - Privacy/compliance depends on your Google Cloud configuration; sending sensitive audio to a third-party cloud API may be inappropriate for regulated data without proper controls. **When to prefer alternatives:** - If you require offline/on-device transcription, or strict data locality, an on-prem STT engine or self-hosted model would be better. - If you need rich metadata (speaker diarization, word timestamps, detailed confidence) or highly specialized language/domain models, a more feature-heavy STT solution might be necessary.