Your interviewer can’t see this. sotto can hear them.
A real-time AI copilot for technical interviews. Invisible to Zoom screen-share, prompted with your CV, and it speaks every language you do.
- Excluded from screen captureWDA_EXCLUDEFROMCAPTURE
- Hear one language, answer anotherCtrl+Shift+L
- Recording governs the transcript, not the interviewPer-profile · new profiles record
Five things that have to work together.
Any one of these on its own is a demo. The interview is the part where all five have to hold at once.
- 01
Invisible
Stealth-firstWindows display affinity excludes the overlay from screen-capture APIs, so Zoom, Teams and Meet record an empty desktop. Checked against OBS Display Capture.
WDA_EXCLUDEFROMCAPTURE - 02
Cross-language
Listen X · answer YPolish recruiter, English answer. Ukrainian tech lead, Ukrainian answer. Set it per profile and switch mid-interview.
Ctrl+Shift+L - 03
Personalized
Your CV, your promptsDrop in your CV, paste the job description, write your own system prompt. Signed in, those files are uploaded to your account — full prompt control, with knowledge-base attachments.
schema_version: 1 - 04
Your model
Pick the brainClaude Sonnet 5.5 by default, Opus when a question deserves more thinking. Each profile names a fallback that takes over if the primary fails mid-answer.
Per-profile fallback - 05
Yours afterwards
Recorded per profileTurn recording on for a profile and that interview syncs to your account — the transcript turn by turn, searchable, and exportable as Markdown or JSON. Leave it off and the interview still appears in your history, without the transcript.
Per-profile recording
One copilot. Every round.
Sotto picks the prompt template that matches the shape of the question — or you pin it from the overlay with a hotkey.
- CodingLeetCode-style. C# / Python / Go / TS.
- System DesignMulti-tenant, scale, trade-offs.
- Object-Oriented DesignParking lot, elevator, chess clock.
- BehavioralSTAR. Tell me about a time…
Three steps. Five minutes.
Install the Windows app
Run it and the app installs for your user only, so no admin rights and no install location to choose. It is 82 MB because the .NET runtime ships inside it, so there is nothing else to install. Later versions install from Settings, when you ask for them. Windows will warn you the publisher is unknown — the build is not code-signed yet.
v0.8.0 · 82 MBConfigure a profile
Drag in your CV, paste the job description, pick a model, and set the language you will be asked in and the language you want back.
schema_version: 1Open Zoom. Press Ctrl+Shift+A.
Sotto streams system audio and your microphone to its transcription provider as separate channels, so it always knows who asked what. The interviewer sees an empty desktop. You read the answer.
Ctrl+Shift+A
The questions you actually have.
Is this legal?
It depends on the process you are in. Sotto is a tool. Some employers prohibit AI assistance during interviews, others encourage research — read what you agreed to, and make your own call.
Will Zoom, Teams or Meet see the overlay?
No. Windows display affinity (WDA_EXCLUDEFROMCAPTURE) excludes the overlay window from the desktop-duplication APIs those apps use, so a screen-share or recording captures an empty desktop. We check it against OBS Display Capture.
Is there a Mac version?
Not today — sotto is Windows only, because the stealth behaviour is built on Windows display affinity. There is no macOS release scheduled.
Do you store my CV or my conversations?
Yes, once you sign in. Your profiles sync to your account — the CV and knowledge-base files you attach are uploaded and stored, along with the text extracted from them — and every interview you run leaves a record there: the profile it used, when it ran, and how many answers it produced. The turn-by-turn transcript is kept as well when the profile has conversation recording on, which is how a new profile is created.
What if the model is down mid-interview?
Every profile names a fallback model. If the primary fails while an answer is streaming, sotto retries on the fallback and tells you in the overlay that it switched.
Stop bombing interviews you could’ve passed.
Free to download. Interview audio goes to a third-party transcription provider, prompt and transcript context to a third-party model provider; signed in, your profiles and every interview reach your account.
Download for Windows