Have something to say?
Support Proton Drive Sync
Please add support for syncing notes, audio, and transcripts with Proton Drive 🙏 For users prioritizing privacy and end-to-end encryption over Google Drive or iCloud, having Proton Drive as a native sync/backup option (or via local folder mapping) would keep data secure across devices.
Fixed minimum window width limit
I use one of my screens in portrait mode with 1080×1960 max. resolution. When Vowen app is placed on this screen app is too wide and extends outside of the screen area without possibility for app window width to be further reduced. can minimum window width limit be adjusted to support this screen resolution?
macOS: paste back into the originating WINDOW, not just the originating app
On Windows, Vowen stores the target window handle and reactivates that specific window before pasting. On macOS it stores only the bundle id and reactivates the application. That difference matters for any app that holds several independent working contexts in one process. I run several Claude Code sessions inside one editor. I hold the dictation key in session A, speak, and while transcription is running I switch to session B. Vowen correctly brings the editor to the front, then pastes into session B, because at the bundle-id level A and B are the same target. Switching to a different application works correctly, so the app-level targeting itself is fine. It is only same-app, different-window that goes wrong. This is a request for parity with the Windows behaviour rather than new capability: capture the frontmost window on macOS at record start and restore it before inserting. Environment: Vowen 0.5.8, macOS, pro plan.
Document the vowen:// URL scheme
The app registers a vowen:// URL scheme and has a deep link handler wired up, but the documentation does not describe it. If it supports actions such as starting a notes session or opening a note, that is a simple automation hook that already exists and that nobody can currently use.
Opt-in authenticated access to the local API from another machine
The local HTTP API is bound to 127.0.0.1, so only software on the same Mac can reach it. Anything self hosted elsewhere cannot. A setting to bind to the LAN, or to accept an authenticated connection from a specific host, would let a self hosted automation platform drive Vowen directly. The bearer token in cli/server.json is already the right mechanism, it just is not reachable from anywhere else. Defaulting to loopback is the correct and safe choice. This is a request for an opt-in, off by default.
Webhook or callback when a note or transcription finishes
vowen watch streams new items as NDJSON, but it is poll based and needs a process sitting running to catch anything. A webhook fired on completion would let an automation react to a finished meeting without holding a process open. A useful payload would be the item id, type, title, duration, and whether transcription succeeded. Use case: when a client call finishes, a workflow picks up the transcript, drafts the follow up and files it, with no polling and nothing to babysit.
Notes start -- template silently ignores an unknown template name
Running vowen notes start --template "NoSuchTemplateXYZ" against a name that does not exist prints "Starting notes session..." and "Notes session started (recording)", and records the meeting. It does not warn that the template was not found. The resulting note carries no record of which template was applied, so there is no way to tell afterwards either. A typo in the template name therefore costs the whole meeting's formatting, silently, and you only find out when you read the summary. Suggested fix: exit non-zero with "template not found: " before starting the session, and list the available template names. Failing that, record the template on the note so it can at least be seen after the fact. Worth noting alongside this: template names are title-cased on save, so a template created as "Client discovery call" is stored as "Client Discovery Call". That makes an exact-match flag easy to get wrong. Environment: Vowen 0.5.8, macOS, pro plan.
Notes regenerate -- transcript empties the note and reports success
On 17 September 2026, running vowen notes regenerate --transcript --diarize --num-speakers 2 against a 50 minute meeting note printed "Transcribing... 0s" then "Regenerated note ", and exited 0. It completed in under twenty seconds, which is not long enough to transcribe fifty minutes of audio. Afterwards the note's transcript, summary and text fields were all empty. errorState was false and processingState was false. Nothing reported a problem. The source audio was untouched and still plays. Re-running produced the same result: instant apparent success, empty transcript. There are two separate issues: Regeneration clears the existing transcript before it has a replacement, so a failure leaves the note empty rather than unchanged. The failure is silent. It reports success, sets no error state, and exits 0. Suggested fix: write the new transcript to a temporary field and only swap it in once transcription has actually produced output, and return a non-zero exit code with an error when it has not. Related: vowen notes update has no --transcript flag, so there is no way to restore a backed-up transcript into a note once this has happened. Environment: Vowen 0.5.8, macOS, whisper large-v3-turbo, pro plan.
Read Aloud shortcut for selexted text
can I select text a read and use a shortcut to start read aloud
AI Chat with all notes
It would be great to have a functionality to chat with all of the content generated with transcription. currently I am only able to chat with individual meeting transcripts. one central chat location would be ideal where it can reference all of the chats.
AI model / voice model backup
Hi Team Vowen, It would be nice to have the possibility to select a backup voice model and AI model in case our primary model selections fail to to do the job. For example, for a voice model, I select my primary model as Gemini 3.5-Transcribe and my secondary/backup model as Whisper Large V3 Turbo. So, if for any reason, Gemini 3.5-transcribe fails to transcribe, then Vowen automatically switches to V3 Turbo to proceed with the transcription. Same thing for AI transcription enhancement models. Great app by the way !
Adding tags to the front matter delimiter block
I use AI to auto create tags for notes, and I’ve updated a Meeting Template to include these tags in the copy. Rather than include in the main note summary block, I’d like to include these tags in the front matter delimiter block (between the opening block \—- and \—- tags). I’ve tried multiple different prompts in the Edit Template field, but none add the tags to this first block. How do I do this? Related, if this is not doable, can we at least add the manually added tags (ie those coloured manual tags added to the note summary at the top of the page) to this front matter delimiter block?
Save Audio Recording location
It would be useful to be able to specify the storage location of saved audio recordings. I’m on OSX, and the recordings are automatically stored in ~/Documents/Vowen Recordings. This was fine until I realised it’s using up all my iCloud storage. I’d prefer to automatically store these files on my local drive or a NAS drive, rather than having to do file maintenance from time to time… This isn’t mission critical, but it would be great to have a set it and forget it option.
Google Calendar Integration and Automatic Recording
I'd like Vowen to connect to Google Calendar so that when a scheduled meeting starts, Vowen starts recording it. At the moment every recording depends on me remembering to open the app and press start. If I forget, the meeting is gone. There's no going back for it afterwards. This is the main thing I'm missing from Granola. The meeting sits in my calendar, Granola knows it's happening, and the recording starts with the meeting. I don't have to think about it. What I'm asking for: Read-only access to Google Calendar Vowen recognises that a scheduled meeting is starting It begins recording, either automatically or with a one-click prompt if you'd rather not start unattended The note is titled from the calendar event, so I can find it later Most of this already exists. There's meeting detection that triggers when an app starts using the microphone, and it can prompt to take notes. What it can't do is know that a meeting was scheduled, when it was due to start, or what it's called. The calendar connection looks like the missing piece rather than a rebuild. Read-only would be enough. I don't need Vowen writing to my calendar, joining calls or contacting attendees. I'm assessing Granola and Vowen side by side, specifically on meeting note capture and summarisation of transcripts. This is the main gap.
Small Improvement
Hey — found Vowen through a Reddit thread today and have been trying it out. Really liking it so far. One small transcription gap I noticed: when I say something like “August nineteenth,” it outputs “August nineteenth” rather than “August 19th.” Would be great if spoken dates could automatically normalize into their natural written format. Just wanted to flag it since it stood out while using it.
Enter pressed decided per transcription
There's an option to always have enter pressed after the transcription is done. It'd be nice if there were two different shortcuts/paths during transcription that could enable or disable the enter press just because most of the times I want enter to be pressed, but sometimes it's a more sensitive message that I will be revising.
Export transcript as word-level JSON with per-word timestamps
The transcribe tool already exports as text, subtitles, PDF, plain text, Markdown, SRT, and VTT — that covers most of what I need. The one format I'd love to see added is a word-level JSON export, where every single word has its own start/end timestamp, rather than grouping multiple words or sentences into one timestamp block (like SRT/VTT do). Adobe Premiere Pro already supports this kind of export. Having it in Vowel would help in two main cases: YouTube chapter/timestamp generation. Feeding an AI the current subtitle export gives it far fewer timestamp anchors, since each block often spans multiple sentences. A word-level JSON gives the AI enough granularity to place chapter markers exactly when a topic starts — instead of chapters that start or end mid-sentence because the underlying block was too coarse. Motion graphics / animation timing. When converting a design into a motion graphic to sync with a video, having every word's exact timestamp means I know precisely when to trigger each animation, instead of estimating from a block-level subtitle. This is achievable today by exporting from Premiere Pro instead, but since I'm already using Vowel's transcription for the whole session, having this JSON export available directly from Vowel would save a full extra step.
Lowering the sound
Currently, while dictating, we have three options: we can either pause the media, mute the sound completely, or do nothing with it. I think a very useful option would be the ability to lower the sound by a set percentage, for example by 50% during dictation, when we do not want to turn the sounds off completely, but only want them to be quieter.
Raycast Hyper Key compatibility issue on macOS
hen I set a Vowen shortcut to ⌃⌥⇧⌘ + , it works if I press the modifiers physically. However, the same shortcut does not trigger when using Raycast's Hyper Key (Caps Lock → Hyper), even though Vowen records the shortcut correctly and Raycast recognizes it. Could you investigate compatibility with Raycast's Hyper Key?