backchannel
Windows x64
For Windows 10 or later on 64-bit Intel or AMD processors.
- Download
Backchannel-windows-x64-v0.5.1.zip. - Unzip it and run
Backchannel.exe. - If SmartScreen appears, choose More info, then Run anyway.
Release v0.5.1
Tells you what a call cost, and stops writing everything down twice.
The Tokens tab is a cost report now: it opens with what the call cost, every agent carries its own figure, and the live audio bridge - billed by the minute rather than by tokens - appears instead of showing nothing. On calls that capture both your microphone and the meeting's own audio, close to half the transcript was the same speech saved twice under two speaker names; those duplicates are now suppressed as they happen.
Download Backchannel v0.5.1
These unsigned open-source builds are delivered through the authenticated Backchannel portal. An approved Backchannel account is required; no GitHub identity or repository membership is required.
The Windows and Linux bundles include ffmpeg. On macOS, voice calibration and compressed audio imports still need ffmpeg on PATH.
macOS Apple Silicon
For Apple Silicon Macs. Intel macOS builds are not included in this release.
- Download
Backchannel-macos-arm64-v0.5.1.zip. - Unzip it and move
Backchannel.appwherever you keep apps. - Right-click
Backchannel.appand choose Open the first time.
Linux x64
A portable bundle for 64-bit Intel or AMD Linux systems.
- Download
Backchannel-linux-x64-v0.5.1.tar.gz. - Extract it with
tar -xzf Backchannel-linux-x64-v0.5.1.tar.gz. - Run
Backchannel/Backchannel.
What changed
v0.5.1 follows v0.5.0 with a cost report in place of a token count, a transcript that stops repeating itself, and bounded analysis replies.
The Tokens tab is a cost report
It opens with what the call cost rather than a token count, and every agent carries its own dollar figure, so the expensive one is obvious at a glance. The live audio bridge is billed by the minute rather than by tokens, which meant it previously showed nothing at all - on a measured 35-minute call that was roughly a fifth of the bill, entirely invisible. It now appears alongside everything else, and the pricing table shows its per-minute rate instead of a dash.
Calls stop transcribing the same sentence twice
On a call that captures both your microphone and the meeting's own audio, close to half of all transcript entries were the same speech saved twice under two different speaker names. That is also why a two-person call could present as four participants. The duplicates are now suppressed as they happen, so the transcript reads properly, the export is not half padding, and every analysis agent stops re-reading the same words for the length of the call.
Runaway answers are capped
A single analysis call in an earlier session produced sixty-three thousand tokens in one reply, almost all of it discarded. Replies are now bounded and reasoning is budgeted. A reply that legitimately needs more room is given it on a second attempt rather than being cut off mid-sentence.
Lighter while a call is running
Live speaker detection uses substantially less processor time, and the call screen no longer redraws itself sixty times a second to animate the audio meter. Long meetings leave the machine responsive.
The meeting-audio option explains itself
The pre-call checkbox now says plainly that your microphone is always recorded, and that the option adds a second track carrying the meeting's own audio - the separation that lets speaker identification tell local from remote voices definitively. Turning it off states the consequence: remote participants are transcribed only if your microphone can hear them, so with headphones the far end is missed. The recording consent notice is legible in dark mode rather than gray-on-gray.
New models, and prices that are correct
GPT Live Transcribe and Gemini 3.7 Flash are selectable. The Gemini Flash rates shown in the pricing table were twice what they should have been and have been corrected, so cost estimates for those models were previously overstated.
Insights stop quietly disappearing
An insight could be deleted by being merged into itself, with no record that it had existed. That can no longer happen. Sales-opportunity scouting is also switched off entirely on meeting types that do not use it, rather than running and discarding the result.
First-run notes
| Need | What to do |
|---|---|
| Release access | Use the email and password for your approved Backchannel recipient account. |
| Provider key | Follow the guided setup on first launch, or add a key later in Admin -> Connections. Calls will not start until the transcription model has a usable credential; local ONNX transcription needs none. |
| Voice calibration | Works out of the box on Windows and Linux. On macOS, install ffmpeg and make sure it is on PATH before enrolling your voice. |
| Upgrading | From this release the app updates itself in place: it verifies the new build's signature first and waits for recording and post-processing to finish before restarting. If you install a build manually instead, quit the old version from the tray first. Session data is stored per-user and shared by every version. A previously configured OpenAI-compatible endpoint is migrated automatically on first launch. The Admin About tab shows a what's-new notice after upgrading. |
| Desktop launch | Backchannel opens a dedicated local app window in your default browser when it supports app windows, otherwise it uses a standard browser tab. It remains available from the system tray or menu bar. |
| Docker | Check out the v0.5.1 source tag and rebuild Compose to run the same release through containers. |
Start with the desktop app or self-host the stack
Both paths run Backchannel on your machine. Choose the desktop bundle for the fastest start, or Compose for maximum control.