What speech formats are evidenced?
Conversations, interviews and speech recordings are part of BBI’s current public capability. The code-switching project used one-hour interviews rather than short isolated utterances.
Capability / Speech and audio
Speech and audio data collection for AI.
Custom conversations, interviews and speech recordings produced around the people, languages and format specified by the project.
The work
Speech production around a real brief.
BBI Media produces original human speech and audio recordings for AI-training data projects.
The current public record includes conversations, long-form interviews, multilingual speech and code-switching. The collection is designed around the required participants and recording format rather than selected from a pre-built catalogue.
Projects can combine participant coordination with the production itself. In the published 2026 code-switching project, BBI handled recruitment, scheduling, recording and delivery for 1,200 participants across eight languages.
This page describes collection and production work BBI can evidence. It does not claim annotation, model evaluation, a proprietary crowd platform or an off-the-shelf dataset library.
Production path
From people to recorded material.
Every collection has its own language, speaker and recording constraints.
- Define the collectionConfirm the speaking format, participant profile, language mix, duration, volume and delivery requirement.
- Recruit participantsSource participants against the agreed profile and project geography.
- Schedule sessionsCoordinate people and recording availability for the planned volume.
- Record speechProduce the agreed conversations, interviews or speech sessions.
- Complete deliveryFinish the production against the agreed brief and hand over the resulting files.
Project evidence
Long-form, multilingual and delivered.
A 2026 code-switching collection shows the operational scale behind the service.
600+one-hour interviews
1,200participants recruited and scheduled
8languages within the project
Full project note
Code-switching speech collection.
Read the exact remit, supplied inputs and delivered scale without inflated claims. ↗
Questions
What buyers need to establish.
Can BBI recruit and schedule speakers?
Yes. Recruitment and scheduling were both within BBI’s remit for the 1,200-participant project. The participant profile and geography still need to be defined for each new brief.
Can a project include several languages?
Yes. The published collection covered eight languages. BBI also has separate localisation experience in 70+ languages, but that does not mean every collection automatically supports all 70+.
Does BBI sell ready-made speech datasets?
This service is presented as custom production. Contact BBI with the required language, participant profile, recording format and volume so the collection can be scoped accurately.
Define the voices you need.
Send the language mix, participant profile, recording format, expected volume and timing.
przemek@bbionline.pl ↗