1h ago
Auto select mode based on recording duration
Right now each mode is chosen manually, but it would be great if Superwhisper could auto-select a mode based on how long the recording is. For example, short recordings (under ~5–10 seconds) could route to a fast, lightweight pipeline — something like Whisper for speech-to-text and GPT-4o Mini for the LLM rewrite. Longer recordings (over ~10–30 seconds) could route to a more powerful pipeline, such as ElevenLabs Scribe v2 for transcription and Claude Sonnet for the rewrite.
This would let users get near-instant results for quick notes/commands while still getting higher-quality transcription and rewriting for longer, more complex dictation without having to manually switch modes each time. Ideally the duration thresholds would be configurable per mode, so users can tune the cutoffs to their own workflow.
Pending