switch transcription model from Whisper to parakeet (faster + better) #28
Labels
No labels
prio_critical
prio_low
type_bug
type_contact
type_issue
type_lead
type_question
type_story
type_task
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
lhumina_code/hero_voice#28
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Overview
hero_voicecurrently uses Whisper for STT. Per Scott's component work, parakeet models are faster and more accurate. Migrate the default STT path to parakeet, keep Whisper as a fallback option.Why
Meeting 2026-05-06 (Scott): "hero_voice component / can use parakeet models / much better than whisper, faster".
Acceptance
Related
Source: meeting notes 2026-05-06.
Migration to Parakeet is complete. For now we use Parakeet v2, which is English only. In the future we can also support Parakeet v3 for European language support (with slightly reduced English accuracy). Our backend library Sherpa ONNX also supports a variety of models, including Whisper, if broader language support is needed in the future.