Transcription Taking Too Long
Understanding why your transcriptions might be slow and how to fix it.
Understanding Transcription Delays
If you find that transcription is slower than expected, there are a few factors that could be contributing to the delay.
Local AI Models & Core ML Optimization
VoiceInk uses local AI models for transcription. For optimal performance on Apple Silicon devices (M1, M2, M3, etc.), VoiceInk utilizes Core ML, Apple's machine learning framework.
When you first download a non-quantized model, VoiceInk performs a one-time optimization process to prepare the model for Core ML. This can take several minutes, depending on your system's resources. Subsequent transcriptions using that model will be significantly faster, often 2-4x faster than without Core ML.
If you are using an Intel-based Mac, local models are almost unusable. For the best UX, it is strongly recommended to use a cloud-based transcription service like Groq or Deepgram. You can configure these in the AI Models section of the app.
AI Enhancement
If you have AI Enhancement enabled, this will add an extra step to the process, which can increase the time it takes to get your final text. The speed of this enhancement depends on the AI provider you are using.
- Gemini: While powerful, Gemini can sometimes be slower, especially on the free tier.
- Groq: Known for its very fast performance.
Performance Analysis
If you want to understand exactly how long each part of the transcription process is taking, you can use the built-in performance analysis tool.
- Go to the Transcription History tab.
- Select the transcription you want to analyze.
- At the bottom of the view, click the Analyze button.
This will show you a breakdown of the time taken for each step, including:
- Transcription Time: The time it took for the AI model to transcribe the audio.
- Enhancement Time: The time it took for the AI enhancement to process the text (if enabled).