AI conversation intelligence
Oralyvo™
Make a conversation easier to find, understand and use.
Oralyvo keeps a conversation in one searchable workspace, so a meeting, lecture or interview does not disappear into a recording nobody wants to scrub through later.
Visit Oralyvo (opens in a new tab)- Ownership
- A WianTribe product
- Year
- 2026
- Platform
- Web · Installable PWA
- Status
- Live
Overview
- Our role
- Product strategy
- Product design
- Full-stack engineering
- AI integration
- Disciplines
- AI
- Web Application
- Product Development
A person can capture live speech or upload a recording. Oralyvo turns it into a timestamped transcript with speaker labels, then lets them create summaries and meeting notes, translate the conversation, ask questions and pull out decisions or next steps.
Oralyvo is a WianTribe product, built end to end by us: audio capture, transcription, AI processing, accounts, storage and the interface people use to work with the conversation afterwards.
The problem
Important things get buried in conversations.
Meetings and interviews are full of decisions, promises and useful ideas. A recording technically preserves all of that, but finding one moment later is slow, and handwritten notes rarely capture everything.
The technical side is awkward too. Browsers handle audio differently, phones interrupt sessions and internet connections drop. If a transcription product quietly loses part of the conversation, people stop trusting it very quickly.
How we handled it
Keep everything tied to the transcript.
The transcript is the centre of the product. Summaries, translation, questions and notes all work from that same source, so people do not have to upload the same conversation to a collection of separate tools.
A dropped connection should not ruin the session
Oralyvo has multiple transcription options, automatic fallbacks and local recovery so a temporary connection problem does not have to mean starting over.
Provider keys stay off the browser
AI provider credentials live on the server. The browser talks to Oralyvo's own endpoints instead of carrying those secrets itself.
Answers should come from what was actually said
Questions about a conversation are answered using the transcript as the source, rather than treating the model's memory as the record of the meeting.
Oralyvo in pictures
Key capabilities
What you can do with one transcript.
- 01
Transcribe live audio
Capture a microphone, a shared browser tab or system audio, with timestamps and speaker labels that can be corrected afterwards.
- 02
See who said what
The Live AI engine separates speakers automatically, making a long transcript much easier to follow.
- 03
Translate as the conversation develops
Finalized phrases are translated beside the original and recent context is carried forward so short follow-up sentences make more sense.
- 04
Turn it into notes and next steps
Create meeting notes, summaries and action items from the transcript without moving the conversation into another app.
- 05
Ask a question about what was said
Ask about the conversation and get an answer grounded in the transcript itself.
- 06
Find it later or take it with you
Saved transcripts are searchable and can be exported to TXT, PDF or DOCX. Study tools can also turn them into quizzes and flashcards.
Engineering
What keeps a live session from falling apart.
Most of the reliability work happens where the person using Oralyvo never sees it.
- 01
Replay the audio that never made it through
Live audio is numbered and held briefly on the device. If the connection drops, Oralyvo opens a new one, sends the pending audio in order and continues the same transcript.
- 02
More than one way to transcribe
Deepgram Nova-3 on Cloudflare Workers AI handles real-time recognition with speaker separation. Browser speech recognition and Groq Whisper are also available, with automatic fallback when an engine is not configured.
- 03
Short-lived access instead of secrets in URLs
AI processing runs through a Cloudflare Worker. Live connections use short-lived, single-use tickets issued through Durable Objects instead of putting account tokens into connection URLs.
- 04
Keep enough locally to recover
Finalized transcript segments are checkpointed on the device, so a session can be recovered after a reload. Active mobile recordings also keep the screen awake where the device supports it.
Technology
Built with
- Cloudflare Workers
- Durable Objects
- Workers AI · Deepgram Nova-3
- Groq Whisper
- OpenAI, Anthropic & Gemini
- Supabase Auth
- IndexedDB
- Progressive Web App
Need help with something similar?
These are the WianTribe services most closely connected to the work shown in Oralyvo.
- Web Application Development
The kind of browser-based products people sign into and use: SaaS, portals, dashboards and internal tools.
- AI Development & Automation
AI features with a clear job to do — from speech and translation to document workflows, summaries and useful automation.
Let’s talk
Got something you're trying to build?
You do not need a finished brief. Tell us what you are working on, what is not working yet and what you would like the product to do. We can take it from there.