Release Notes - v0.9.6-alpha

A feature release centered on merging recordings, with new transcript export options and a large internal consolidation of how every ingestion path resolves its transcription settings. It is backwards compatible with the v0.9.x line, and all database changes run automatically on startup.

New Features

• Merge recordings (#323). Several recordings can now be combined into a single recording that is re-processed from scratch. The most common need is a dropped call or an interrupted recording that had to be restarted, leaving two partial transcripts and two half summaries where there should be one. Merging concatenates the source audio, in the order you choose, and runs the combined file through the full pipeline: transcription, diarization, summary, and automatic speaker labelling, exactly as a normal upload would. There are two ways to start a merge:

  • From the sidebar. Enter selection mode, tick two or more recordings, and choose Merge. A dialog lets you reorder the sources, add more from a searchable list, set a title, and optionally delete the originals once the merged recording is created.
  • From the recording view. After finishing an in-app recording, the "Upload Recording & Notes" button is now a split button whose menu offers "Merge with an existing recording." You pick which existing recording(s) this clip continues, and the clip is appended to them and merged directly, without being transcribed on its own first.

• Choose which notes and metadata carry over on a merge. Notes cannot be meaningfully concatenated, so when more than one source has notes the merge dialog lets you pick which recording's notes to keep (or keep none). That same reference recording's prompt-template variables carry over as well, so the merged summary uses the intended context. Participants and tags are additive, so they are combined across all sources rather than lost. Folder membership follows the first source.

• Export transcripts as Markdown (#322). The transcript download menu gains a compact TXT / MD toggle. Choosing MD downloads the transcript as a .md file with a title heading, using whichever template you pick, and the choice is remembered.

• Export existing recordings to disk (#321). When automatic export is enabled, Settings gains an "Export all to disk" button that writes every one of your already-processed recordings to the export directory in one step. This backfills recordings created before automatic export was turned on, using the same format and per-user layout as the automatic path.

Improvements

• One shared resolver for transcription settings across every ingestion path. Uploads, reprocessing, merges, recording-session finalization, the share target, and the auto-process folder now resolve language, speaker-count hints, hotwords, the initial prompt, and the transcription model through a single precedence chain (per-request value, then tag, folder, environment, account, and admin defaults). Previously several of these paths each built their own partial parameter set, so a recording created by, for example, the merge or auto-process path could silently skip the transcription model or tag and folder defaults and therefore transcribe differently. All paths now behave identically to a standard upload, which is what makes automatic speaker labelling work on merged recordings.

• Empty language falls through to defaults everywhere. An unset language on any path now resolves through the tag, folder, and account chain rather than being treated as a hard "auto-detect", and folder defaults fill any gaps a tag left open. Automatic detection still results when nothing in the chain sets a language.

• The merge recording picker searches server-side. The dialog for adding recordings to a merge pages results from the server and searches there, so it never downloads the entire recording list, matching how the sidebar already works.

• The recording list updates live after a merge. When a merge completes, the merged recording appears and any deleted source recordings drop out of the sidebar on their own, without a manual refresh; an open recording that was consumed by the merge is cleared.

Bug Fixes

• Merged-recording playback is disabled until its audio is ready. A merged recording is created before its audio has been assembled, so its play button now stays disabled, with a short "audio is still being prepared" note, until the file is available. This also stops an uncaught media error in the console.

• Readable TXT / MD toggle in dark mode. The selected side of the transcript-format toggle used the light accent color as a fill, which left white text hard to read in dark themes. It now uses the saturated button color, so it is legible in every theme.

Compatibility

This release is backwards compatible with the v0.9.x line. Existing recordings, transcripts, and settings are unaffected, and database migrations run automatically on startup. Merging is available to all users; the "Export all to disk" button appears only when automatic export is enabled on the server.