BastionGPT's AI medical scribe is built to capture both sides of a telehealth encounter, giving clinicians a complete record of the visit without any extra equipment or complicated workflows. Whether you're running a quick follow-up or a full 4-hour long intake session, the platform is designed to transcribe the full clinical conversation securely.
The simplest approach is to keep your computer speakers turned on and skip the headphones during the visit. Your microphone will naturally pick up the patient's voice coming through the speakers along with your own, producing a clean two-sided transcript. For best results, set the speaker volume to roughly match your normal speaking voice. This keeps both audio sources balanced and helps the HIPAA compliant AI assistant generate progress notes that reflect the full exchange rather than just the provider's side.
A small percentage of microphones, roughly five percent, don't handle this speaker-pickup method well due to aggressive echo cancellation or noise suppression built into the hardware. If you find your transcripts are only capturing your voice, a reliable workaround is to place a phone or tablet running BastionGPT midway between you and your computer speakers. Positioned this way, the secondary device captures both voices evenly and feeds the session into the same secure healthcare AI platform you already trust for clinical documentation.
If you prefer to wear headphones during telehealth visits, perhaps for privacy in a shared workspace or better audio clarity on your end, BastionGPT still has you covered. Just select the "include audio from a virtual meeting" option before starting your session. This routes the audio directly from your video conferencing software into the scribe, capturing both participants cleanly regardless of your headphone setup. It works across the common telehealth and video meeting tools clinicians rely on day to day.
Does BastionGPT work with Zoom and other meeting platforms?
Yes. BastionGPT works with any telehealth or meeting platform, including Zoom, because it captures audio directly from your device (your microphone and speakers) rather than connecting to the meeting itself. That design has two practical benefits:
Your patients join the visit exactly as they always do, whether that's the Zoom desktop client, a web browser, or a phone. They don't need to install or change anything.
No bot appears as an extra "participant" on the call. Our clinical advisory board evaluated scribes that join meetings as a silent attendee, and patient feedback was consistent: an extra presence during a personal conversation feels off-putting. Capturing audio through your own device keeps the visit feeling like a normal session.
Do I need to install anything?
No. BastionGPT runs entirely in your web browser, and Safari, Firefox, Chrome, and Edge are all supported. There is no desktop app to run and no browser extension to manage. Every transcription feature available on desktop also works on mobile, so some clinicians simply place a phone between themselves and the client and record from there.
Can I upload a recording I already have?
Yes. Use the import file option to upload an existing audio or video recording, such as a session captured on a dedicated recorder or a recorded video visit. We support common audio formats (MP3, WMA, and others), files up to 800 MB, and sessions up to 4 hours long. If a video file is too large, converting it to audio-only usually brings it under the limit.
How does the scribe know who said what?
BastionGPT separates speakers by analyzing the voice prints in your audio, not by guessing from the words alone. Each voice is recognized and automatically labeled by role, such as therapist and patient, and in group sessions the AI picks up names as participants introduce themselves. Voice-print analysis is what keeps labeling reliable in situations that confuse word-based transcription, like two people sharing one camera or a parent and child in the same room. You can also select the dialect that best fits the session, including Canadian English, which improves accuracy for accents while still recognizing clinical terminology.
For a deeper look at the technology behind this, read about why speaker recognition matters in scribes.
How long does my transcript take?
Plan on 5 to 60 minutes, depending on the length of the session. The wait is a deliberate design choice: we analyze every voice across the full recording to label speakers as accurately as possible, because in clinical documentation accuracy matters more than speed. A quicker transcription option is planned for workflows where an instant draft matters more than the deeper analysis.
However you choose to capture the session, every recording and transcript is processed inside a HIPAA compliant environment purpose-built for healthcare. BastionGPT stands apart as a leading AI scribe for doctors and therapists precisely because it combines flexible capture options with the privacy safeguards that clinical work demands, so you can focus on the patient in front of you while the documentation takes care of itself.
Once your note is ready, see getting scribe notes into your EHR for the next step in your workflow.