Collect high-quality, multilingual voice data for TTS and STT. Simple, fast, and reliable.
PDF & Text
Turn into recording clips
Guided Studio
4–15s auto-save takes
Export Ready
Train TTS & STT models
10K+
Contributors
∞Open
Extensible Dialects
100%
Open Data Options
Built for
Researchers & Builders
Let's make voices matter.
3
Active Projects
142
Completed Clips
28
Remaining
2.4 h
Total Recorded
Kurdish Common Texts
2,500 clips · ckb
Arabic News Dataset
1,800 clips · ar
New recording submitted
2 minutes ago
Clip approved
12 minutes ago
Real People
Real Voices
A Better Tomorrow.
See QAI Voice in action
Five short guides show how a project moves through collection, recording, review, and export.
Workspace Preview
Test drive the self-hosted command center. Click between navigation tabs inside the dashboard preview to inspect real-time telemetry, dataset projects, audio verification, and exporter tools.
120 / 500 hours
Here's what's happening with your voice datasets.
12
8,421
6,973
48.2
New recording submitted
2 minutes ago
Clip approved
12 minutes ago
New speaker joined
1 hour ago
From low-resource dialect preservation to high-volume commercial TTS dataset generation, QAI Voice delivers verifiable audio.
Produce clean 24kHz studio datasets with phoneme-aligned transcripts for training XTTS, VITS, and FastSpeech acoustic models.
Collect thousands of varied acoustic takes across Sorani & Kurmanji Kurdish, Arabic dialects, and accented English for Whisper fine-tuning.
Safeguard endangered oral heritage and regional languages with structured community recording pipelines and full speaker consent.
Export ethical, license-compliant audio corpora with rich metadata, verifiable SNR acoustic stats, and standardized train/dev/test splits.
Test drive our studio recording engine. Toggle between raw microphone takes and normalized audio inspected by real-time acoustic quality gates.
Sorani (Central) · Classical Literature & Lexicon
ئەمڕۆ کەشوهەوا زۆر جوان و سازگارە، و دەمەوێت زمانی شیرینی کوردی بە جوانی فێرببم.
“Today the weather is very pleasant and refreshing, and I want to learn the sweet Kurdish language gracefully.”
Target: -16.0 ± 0.5 LUFS
Studio Reference
0 Clipped Samples
24.0 kHz
16-bit PCM · Mono (1.0)
WAV PCM Uncompressed
100% self-hosted on your own hardware or deployed on managed distributed GPU clusters.
Deploy QAI Voice in minutes on your local workstations or on-premise servers with Docker Compose.
Designed for large research institutes, language authorities, and commercial voice AI laboratories.