Tuned for the room
Dictation is handled with Nigerian English and obstetric terminology in mind, not generic office speech.
Built for Nigerian radiology departments
Dictate the scan.Walk away with the report.
SonoScribe listens while you scan, transcribes obstetric dictation, pulls out the measurements, and hands you an editable draft report in seconds — so the reporting happens at the couch, not after the list.
Download for Windows
v1.2.0 · 22.1 MB · Windows 10/11 64-bit
Open the web app
No install · Android APK
- Windows desktop
- Android
- Web browser
- Works offline
10–30%
Target reduction in reporting time
>80%
Transcription accuracy target
3
Obstetric report templates built in
0
Patient records kept on our servers
Why we built it
Reporting is still the slowest part of the scan.
In many Nigerian hospitals ultrasound reporting is done by hand, after the patient has left the room. In high-volume centres that means delay, fatigue and reports that read differently depending on who typed them.
- Findings written on paper, then retyped into a template later.
- Reporting fatigue at the end of a long list, when errors creep in.
- No shared format, so two sonographers produce two different reports.
- General-purpose dictation tools mishear Nigerian English and obstetric terminology.
What changes
The report is written while you scan.
- Speak naturally in the room — no keyboard, no pausing the study.
- Measurements land in the right field of the right template automatically.
- Every report follows the same structure, whoever performed the scan.
- You review, correct and sign off — the draft never leaves your hands.
How it works
Four steps, one sitting.
From selecting the patient to a signed PDF, without leaving the scanning room.
-
1
Pick patient & scan
Choose Early OB – MSD, Early OB – CRL or a full Obstetric Scan. That choice loads the matching report template and its checklist.
-
2
Record while you scan
Hit record and dictate as you work. The template checklist stays on screen so you can see what you have and haven't covered.
-
3
Transcribe & extract
Audio is transcribed, then a hybrid AI and rules pass pulls out MSD, CRL, gestational age, EDD, FHR, biometry and the rest.
-
4
Review & export
Check the structured form, fix anything flagged for review, edit the narrative, then export or print the PDF.
Features
Built around how obstetric scans are actually reported.
Real measurements, extracted
MSD, CRL, gestational age, EDD, FHR, BPD/HC/AC/FL, EFW, placenta, AFI/DVP and cervical length.
Templates that match your forms
MSD stage, CRL stage and second-trimester obstetric templates, structured section by section.
Works when the network doesn't
Recordings are captured on device with on-device transcription as a fallback, then queued and uploaded when you reconnect.
Editable, with history
Every draft stays editable and each revision is kept, so you can see what changed before you finalize.
PDF export & print
Finalize to a clean, branded PDF you can print at the desk or hand to the patient.
Patients and recordings, on device
Patient records, audio and finalized reports are cached locally so yesterday's list is still there without a connection.
It never invents a number
Anything you didn't dictate is left blank and flagged Requires Review rather than filled in with a guess.
Get SonoScribe
Install it on the machine beside the scanner.
Free to download. Windows is the primary target; Android and the browser cover the rest.
Windows
The full desktop app, including on-device transcription for when the network drops. Windows 10 or 11.
Download installerHeads up: the installer isn't code-signed yet, so Windows SmartScreen may warn you. Choose More info → Run anyway to continue.
Android
Same app on a phone or tablet, for scanning rooms without a dedicated PC.
Download APK- Download the APK on the device — it's a large file, so use Wi-Fi.
- Allow installs from this source when Android asks.
- Open the file and install.
Not on Play Store. Installing an APK directly is expected here — the download link on this page hands off to our GitHub release page over HTTPS.
Web
Open it in a browser on any machine — useful for a quick look, a shared workstation, or a demo.
Open app.sonoscribe.africaOne difference: on-device offline transcription isn't available in the browser, so the web app needs a connection to transcribe.
macOS and Linux aren't packaged yet, but the app builds for both from source.
SonoScribe writes drafts. Clinicians write reports.
Everything SonoScribe produces is a draft for review and sign-off by a qualified clinician. It does not diagnose. It does not invent measurements you did not dictate — fields it cannot fill are left blank and marked Requires Review, and studies it is unsure about are flagged rather than quietly completed.
Patient-identifiable data is not persisted on the server. Recordings, transcripts, patient records and finalized reports stay on the device that made them; the backend keeps only redacted workflow audit records.
SonoScribe is a research prototype developed for evaluation in Nigerian diagnostic settings, not a certified medical device.
For developers
There's an API behind all of this.
The reporting pipeline — transcription, clinical extraction, template selection and
draft generation — is a FastAPI service. Authenticate with an X-API-Key
header and post audio or a transcript. Access is issued per deployment; the schema
isn't published openly.
# audio in, draft report out POST /api/v1/pipeline/audio-to-report Host: www.sonoscribe.africa X-API-Key: <your key> file=@dictation.m4a scan_type_hint=early_obstetric