Catalan conversational speech dataset
Spirelight lists a conversational speech dataset for Catalan with 500 declared audio hours, 50 listed speakers and Spain as the recording country, at $75 per audio hour. Request a sample to assess recording and transcript coverage; confirm provenance, current availability and written terms for your intended use with our team.
Need another dialect, speaker mix, device, recording condition or rights scope? Commission speech data to your specification. This listing describes an existing dataset; a new collection is quoted separately.
The video couldn’t start. Try playing it again or open the video directly.
Dataset specifications
| Language | Catalan |
|---|---|
| Recording country (listed) | Spain |
| Speech type (listed) | Conversational; inspect the sample’s speaking style |
| Declared audio volume | 500 hours; confirm the available selection |
| Listed speaker count | 50; request the speaker distribution and inclusion criteria |
| Audio | Confirm sample rate, file format and channel layout |
| Transcripts | Confirm coverage, timestamps, conventions and review method |
| Metadata | Request field definitions and coverage for the selected recordings |
| Rights | Confirm provenance, consent documentation and permitted uses in the asset-specific written agreement. Licence review checklist |
| Listed price | $75 per audio hour; confirm selection, deliverables and total in the quote |
Listing metadata is dated 2026-10-12; this is not a current-stock or delivery confirmation. If the available recordings do not cover your brief, ask us to assess custom speech data collection feasibility and scope.
Need a buying checklist? Our guide to buying AI training data explains how to compare a candidate sample, documentation, acceptance checks and written terms.
Evaluating the listed Catalan dataset for a purchase? Request a written quote for your selected recordings, available hours, deliverables, permitted uses and total price.
Assess against your task
A language label and hour count do not establish that a dataset fits your model. Use a sample and supporting documentation to check:
- ASR / speech-to-text: speaking style, recording conditions and reference transcript conventions.
- Speaker diarization: channel layout, speaker-turn labels and overlap references.
- Voice agents: whether accents, turn-taking and acoustics resemble your target conversations.
- TTS or voice synthesis: whether the recordings, speaker permissions and written terms support that separate use.
Does conversation fit your task?
Read speech and conversation differ in turn-taking, hesitation and overlap. Compare the actual sample with your target audio before deciding; the listed language and volume alone do not prove a model fit. Our conversational speech data guide explains the trade-offs.
Request a sample of the Catalan dataset
Tell us which accents, recording conditions, transcript conventions or metadata fields you need to inspect. The team will assess sample availability and confirm what can be shared, any conditions and timing. A request does not reserve the listed hours or confirm a licence.
Frequently asked questions
What is listed in the Catalan speech dataset?
The published listing declares 500 audio hours, 50 speakers, Spain as the recording country and a price of $75 per audio hour. Confirm selected recordings, current availability, transcript coverage, channel layout and metadata with a sample and supporting documentation.
What should I check when evaluating Catalan speech data?
Catalan vowel pronunciations vary by dialect, including reduced vowels in some unstressed syllables and distinct open vowel sounds. Speech also differs across Barcelona, Valencia and the Balearic Islands. If your users switch between Catalan and Spanish, inspect whether the selected audio includes that pattern and how the reference transcript marks it.
Can I use the Catalan dataset for commercial AI training?
Confirm your intended use against the asset-specific written agreement and supporting provenance and consent documentation before purchase. Commercial ASR training, evaluation and voice synthesis may require different permissions. The speech-data licensing guide is a review checklist, not the agreement for this asset.
What does the Catalan speech dataset cost?
The published rate is $75 per audio hour. Ask for a written quote identifying the recordings or selection, deliverables, usage rights, available hours and total price. Any volume arrangement needs explicit confirmation.
What happens after I request a sample?
The form records your request for the team. We will confirm whether a suitable sample is available, what can be shared and the timing. Request acceptance does not reserve listed hours or grant a licence.