Vocabulary
Terms the model should trust
Add names, brands and jargon once, and the speech model is nudged toward them on every dictation. Validated on entry, so a typo never becomes a permanent bias.
Feature
It learns
your nouns.
Every field has its Priyas, its Kubernetes-es and its product codenames that generic speech models mangle. Velora keeps a personal dictionary that fixes them, and learns new ones from the corrections you make.
How it learns
Corrections in, accuracy out.
Replacements
Always-wrong, auto-fixed
When a term is transcribed wrong the same way every time, a replacement rule rewrites it after the fact: deterministic, visible, editable.
Learning
Your edits teach it
Velora learns from the corrections you actually make and from terms that keep recurring in your dictations, quietly, on your Mac. Every learned term is listed in the Dictionary pane, where you can forget one or all of them. A forgotten auto-learned word is banned from being picked up again; a forgotten correction can return if you keep making the same edit.
Portable
CSV in, CSV out
Bring a term list from anywhere, export yours anytime. Sync uses your own iCloud Drive whenever it is signed in and carries confirmed dictionary terms only: never audio, never transcripts.
The fine print
- The Whisper speech models consume the dictionary during transcription; the Parakeet models do not yet — their replacements still apply after the fact.
- Automatically learned vocabulary is capped at the 100 most recent terms; older ones fall off as new ones arrive. Terms you add yourself have no such cap.
- Sync runs whenever iCloud Drive is signed in. It sends your confirmed terms through your iCloud Drive under Apple’s terms, not through a Velora server. There isn’t one. Without iCloud Drive, the dictionary stays on this Mac.
Free. Open source. Local.
Stop typing. Keep writing.
Already using Velora? A star is how the next Mac user finds it.