Vocalstock by Datailix · for AI teams
Real voices, recorded with consent, licensed by the hour for AI.
Native speakers record solo speech and natural two-person conversations in their own language and dialect. You order the hours you need; we deliver a balanced set across many verified speakers, with optional human-typed transcripts, under a clear commercial license for training and evaluating models.
13
languages & dialects
1
verified speakers on sale
4h
hours available now
Training data you can trust, by the hour
Datailix runs the platform end to end: it verifies every speaker, reviews every recording, and licenses the data to companies under clear terms. Accounts are for companies only and are approved by our team.
The data
Balanced, verified, ready to train on
- Choose a language or dialect, solo speech or two-speaker conversations, audio or video, and how many hours you need.
- Every package is spread across many different speakers, and you never receive the same recording twice.
- Native speakers only, each voice-verified, each recording reviewed by a person.
The license
Clear rights for model training
- Perpetual, worldwide license to train and evaluate models; you own the models you build.
- Optional word-for-word transcripts typed by people, delivered as JSON, SRT and CSV.
- A CSV manifest of every file with anonymous speaker IDs. Read the Data License.
How it works
- Request an accountWith your company email. Our team approves company accounts, usually within one working day.
- Order the hours you needSee hours per language, listen to samples and see prices; we pick a balanced set and email you an invoice.
- Pay and upload your receiptWithin 7 days, by bank transfer or another listed method. Orders over 25 hours are delivered in milestones: 50% before each milestone, the rest after you approve it.
- DownloadOnce payment is confirmed, download every file, its transcript if ordered, and the manifest.
Built on consent and trust
Verified voicesEvery speaker records a verification phrase that a person reviews before they can upload.
Clear consentSpeakers accept a written agreement; conversations need both speakers to confirm.
Private by defaultFiles are never public. You get anonymous speaker IDs, never names.
Protected useThe license forbids impersonating a speaker or cloning their identifiable voice.