Skip to content

fix(crispasr): support long VibeVoice transcripts - #11144

Open
localai-org-maint-bot wants to merge 2 commits into
masterfrom
bot/issue-10701-crispasr-long-form
Open

fix(crispasr): support long VibeVoice transcripts#11144
localai-org-maint-bot wants to merge 2 commits into
masterfrom
bot/issue-10701-crispasr-long-form

Conversation

@localai-org-maint-bot

Copy link
Copy Markdown
Collaborator

Description

Fixes #10701.

Pins CrispASR to the fix from CrispStrobe/CrispASR#315. VibeVoice now scales its automatic generation budget with audio duration instead of truncating every transcript at 512 generated tokens, while explicit session overrides remain exact.

The source fetch now requests the exact pinned revision so the dependency commit is available while the upstream PR is under review.

Notes for Reviewers

  • Upstream PR: fix(vibevoice): scale long-form ASR token budget CrispStrobe/CrispASR#315
  • Verified the pinned SHA can be fetched directly from the canonical repository.
  • Verified the pinned source contains the zero-sentinel automatic budget implementation.
  • make -n -C backend/go/crispasr sources/CrispASR renders the expected exact-revision fetch.
  • git diff --check passes.
  • Full backend build was not run; CMake is unavailable on this host and the repository guidance asks before long backend builds.

Signed commits

  • Yes, I signed my commits.
  • Documentation updated (docs/content/) for user-facing changes, or not applicable — dependency bug fix; no LocalAI-facing API/config change.

Comment thread backend/go/crispasr/Makefile Outdated
@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)

# CrispASR version (release tag)
CRISPASR_REPO?=https://github.com/CrispStrobe/CrispASR

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shouldn't this point to the fork @localai-org-maint-bot ?

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed in 5f3bd22: the backend now fetches the pinned CrispASR commit from localai-org-maint-bot/CrispASR, where that commit exists. The expanded Makefile fetch command and diff checks pass.

@localai-org-maint-bot
localai-org-maint-bot force-pushed the bot/issue-10701-crispasr-long-form branch from 70aab26 to a0a285e Compare July 27, 2026 21:03
Pin CrispASR to the VibeVoice duration-scaled generation budget fix so long diarized transcripts are not truncated at 512 tokens. Fetch the exact pinned revision so the dependent upstream pull request commit is available before merge.

Assisted-by: Codex:gpt-5
The long-form VibeVoice fix is not yet in the upstream repository, so fetch the pinned commit from the maintainer bot fork that owns it.

Assisted-by: Codex:gpt-5
@localai-org-maint-bot
localai-org-maint-bot force-pushed the bot/issue-10701-crispasr-long-form branch from 5f3bd22 to 9dd577d Compare July 27, 2026 22:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Transcription with vibevoice-crispasr is not covering audio full-length

2 participants