Why Is Transcription Running Slowly

Because ScribeForge transcribes locally rather than in the cloud, speed depends on your computer's hardware and a few settings — here's what to check if a transcription is taking longer than expected.

1. You're using a large model on modest hardware

The medium and large Whisper models are significantly slower than tiny, base or small, especially without a GPU. If speed matters more than maximum accuracy for a given file, try a smaller model — see choosing the right model.

2. GPU acceleration is off

If your machine supports it, turning on GPU acceleration (Settings → General — Metal on Mac, Vulkan on Windows) can substantially speed up transcription, particularly with larger models. See enabling GPU acceleration.

3. CPU thread count is set too low

Settings → General → CPU threads controls how many processor cores ScribeForge uses for inference. On a multi-core machine, increasing this can speed things up — leave a couple of cores free for the rest of your system if you're multitasking.

4. Very long files

A multi-hour recording simply takes longer than a 10-minute clip, proportionally. This is expected — there's no artificial slowdown, just more audio to process. Batch processing (Pro) lets you queue long files and work on something else while they process in the background.

5. Other apps competing for resources

Heavy background apps (video editors, other AI tools, browser tabs with lots of video) compete for the same CPU/GPU resources. Closing unnecessary apps during a large transcription job can help.