Back to portfolio
Visit live site
Case Study · 01 · Voice & Language AI

Nitq.ai. Brand and product UI for an Azerbaijani voice AI platform.

Two AI models — speech-to-text and text-to-speech — on a single product surface. Designed so journalists, content teams, and developers can try the models within seconds of arriving on the site.

Year2026 RoleLead UX / UI Designer Duration3 months ScopeBrand · UX · UI
nitq.ai · live Nitq.ai voice AI platform
Overview

A voice AI platform for Azerbaijani.

Azerbaijani had limited production-grade tooling for speech-to-text or text-to-speech. International tools mishandled the alphabet and required reading documentation before trying anything. Nitq.ai launched two models on one product surface, with a working demo as the primary entry point on the homepage.

Scope of work
  • 01Brand identity
  • 02Product UI
  • 03Design system
  • 04Bilingual EN / AZ
Every model opens with a working demo. Marketing copy sits below.
Selected screens

Selected screens from the new platform.

Nitq.ai homepage — hero, platform discovery, and capabilities, top to bottom
Homepage

Homepage

A working transcription field opens the page. Platform discovery and capability cards follow — every model is one scroll away from the headline.

Text-to-speech panel with five Azerbaijani voices
Text-to-speech

Text-to-speech

Pick a voice, type Azerbaijani text, hear it back. Five distinct voices, in-browser playback.

Live stenograph view transcribing speech in real time
Speech-to-text

Live transcription

Speak into the mic; the transcript builds word by word, with accuracy and latency reported live.

Authenticated STT workspace with sidebar and request archive
Workspace

STT workspace

An authenticated dashboard for jobs and archive — upload audio or capture from the mic, return to results when ready.

Capabilities overview: STT, TTS, stenograph, quotas
Capabilities

One platform, four surfaces

Speech-to-text, text-to-speech, real-time stenograph, and quotas — described in plain Azerbaijani, on a shared component set.

Mobile menu in Azerbaijani — hamburger expanded
Localisation

Localisation, end to end

Azerbaijani by default — alphabet, the i / İ rule, casing and validation handled at the system layer, across every viewport.

Approach

Design principles.

01

Try, then read

Both models open with a working demo. Marketing copy sits below it.

02

Locale at the system layer

The alphabet, the i and İ rule, casing, and validation behave correctly by default.

03

Two models, one surface

Speech-to-text and text-to-speech share a single header, footer, and component set.

Process

Process.

  1. 01

    Research

    Nine interviews with journalists, content teams, and developers.

  2. 02

    Audit

    Seven international voice AI products evaluated on the same test set.

  3. 03

    Principles

    Try then read, locale at the system layer, one product surface, and perceived speed first.

  4. 04

    Prototype

    Built the hero as a working transcription field, tested with five users before any visual design.

  5. 05

    Launch

    Live demo, streaming transcript, real-time waveform. A small design system was handed over for the team to extend.

Outcomes

Outcomes.

2
Models on one surface
4
Code sample languages
EN / AZ
Bilingual UI
AA
WCAG 2.2 audit
See it live

Now live at nitq.ai.

Visit nitq.ai Back to all projects