Conversational speech data · 60 language and dialect listings

Speech data for the next generation of voice AI

Scope a custom speech or video collection, or explore existing conversational speech datasets. We confirm project feasibility and, for listed assets, sample fit, rights and availability.

Request a quote Browse datasets
5.0 on Datarade 4.5 on Trustpilot SKI Verified Supplier Backed by
60 Language and dialect listings
49 Distinct language names
73,000h Summed listed audio hours
2,000h Largest listed volume
Choose a route

Build to your specification or license an existing speech dataset.

A commissioned collection is planned around your model, participants, conditions, acceptance rules and rights. The speech catalog lists existing assets with their own published specifications and terms.

Custom video and avatars

Scope recruited, consented video for talking heads, lip sync, audiovisual speech or another defined capture task.

Explore video collection

Custom speech and voice

Specify languages, speaker quotas, recording environments, devices, labels and acceptance criteria for new audio.

Explore speech collection

Licensed speech datasets

Compare existing language listings by published hours, speakers and price. Confirm the selected asset's availability and rights.

Browse the catalog

Building an avatar or visual-speech model? See the talking-head collection and audiovisual speech collection requirements.

Need completed-board images or footage of someone writing? Explore custom whiteboard collection.

Meet the founder

The hard part isn’t the training

Andreas, our CEO, describes how we recruit speakers, check recordings and annotate speech for custom collection projects. For a catalog listing, ask us to confirm its specific contents and terms.

Book a call with Andreas Browse datasets

Catalog clarification: The film says “more than 73,000 hours across 60 different languages.” Our published catalog lists 73,000 audio hours in 60 language and dialect listings, representing 49 distinct language names. Browse the catalog and ask us to confirm current availability or additional capacity for your project.

Featured datasets

Explore conversational speech datasets

Compare published listings by language, hours and speakers. Request a sample to check recording format and transcript coverage. Confirm provenance, rights and current availability with our team for the asset you need.

See all datasets
Why teams buy from us

Check the data and terms before you buy

01

Check speaking style

Use a sample to assess turn-taking, disfluencies, overlap and recording conditions for your task.

02

Review rights and provenance

Ask for the asset-specific permission documents and written licence terms for your intended use.

03

Inspect transcripts and metadata

Confirm coverage, timestamps, review method and available speaker fields for the recordings selected.

04

Samples before you decide

Request a sample from a dataset listing and inspect audio quality, transcript coverage and speaker mix. We will confirm which sample materials are available.

05

Published per-hour prices

Each dataset page shows a listed per-hour price. Request a quote to confirm the selection, deliverables, rights and final terms.

06

Scope missing coverage

If the catalog does not fit, describe your language, dialect or recording needs in a custom collection brief. We will confirm feasibility, schedule, rights and pricing.

Coverage

From high-demand markets to underserved languages

English Spanish French German Italian Portuguese Russian Greek Japanese Korean Mandarin Marathi Turkish Indonesian Vietnamese Swahili Yoruba Amharic All datasets
Buyer guides

Know what to ask before you buy

Buying

How to buy AI training data

Build or buy, how to vet a vendor on consent and QA, and the license terms to check before you sign.

 ASR

ASR training data

What makes speech recognition data accurate: accent coverage, recording conditions, transcript quality and a held-out test set.

 TTS

TTS training data

Clean studio audio, single or multi-speaker design, and the phonetic coverage a natural voice needs.

All buyer guides

Tell us the speech data you need

Send your speaker profiles, languages, and recording conditions. We review the brief and reply with feasibility questions and next steps.

Request a quote