Settings Overview

Settings

Transcription settings overview screenshot

Screenshot

What This Page Solves

Many setup problems are not caused by one bad toggle. They come from changing the right things in the wrong order.

Typical examples:

• downloading large models before choosing the right model path
• turning on GPU before proving the CPU baseline is stable
• building AI prompt templates before the provider connection is working

This page gives you the order that avoids the most rework.

Recommended Setup Order

1. Start with general settings such as paths, language, and theme.
2. Move to transcription defaults and download the baseline models.
3. Configure download behavior if you rely on batches or link workflows.
4. Set up AI services only after the local transcription path is stable.
5. Tune GPU, concurrency, or advanced parameters last.

What To Check First In Each Area

General

• model and cache paths
• language and UI preferences
• whether old-machine migration is needed

Transcription

• default model route: Whisper or realtime model
• default language
• GPU on or off
• whether advanced params should stay at defaults

Downloads

• concurrency
• retry behavior
• proxy or restricted-network settings
• link-transcription toolchain readiness

AI services and prompts

• provider connectivity
• default model selection
• whether prompt templates should be shared or local
• compliance or security boundaries

Common Mistakes And Troubleshooting

• Downloading many models before deciding paths Finalize paths first, then download.
• Maxing GPU and concurrency on first install Build a stable baseline first, then scale up.
• Designing prompt templates before provider health is confirmed Verify connectivity first.
• Focusing only on models, not storage and cache Many apparent model issues are actually path or capacity issues.

Check problems in this order:

1. writable paths and free space
2. model and toolchain completeness
3. GPU or proxy as added variables
4. whether too many settings were changed at once

Read Next

• Device- and task-based model selection: Model Usage Recommendations
• When GPU is worth enabling: GPU Transcription
• Parameter tuning after the baseline is stable: Advanced Parameter Transcription