Voice and speech recordings
Scripted and spontaneous speech, conversational audio and command phrases, recorded across speakers, accents and environments you define.
AI Training Data Collection
Custom AI training data, collected to your specification
Off-the-shelf datasets rarely match the edge cases your model actually fails on. One6 collects new human-generated data — voice, text, images, video and custom formats — designed around your model's target use, demographics and quality bar.
What we deliver
Text and conversational data
Human-written prompts, responses, dialogues and domain documents for fine-tuning and instruction data.
Image and video capture
Photos and video of objects, scenes, gestures and activities, captured to defined framing, lighting and diversity requirements.
Custom collection programs
Bespoke tasks when your data doesn't fit a standard category — we design the collection protocol with you before the first contributor starts.
Common use cases
- Training speech recognition and voice assistants on underrepresented accents
- Building fine-tuning and instruction datasets for LLMs
- Collecting computer vision data for real-world edge cases
- Creating held-out test sets that were never published online
How a project works
- 01
Scope
A short call to understand your model, data requirements, quality bar and timeline.
- 02
Pilot
A small first batch so you can review real output before committing to volume.
- 03
Scale
Once the pilot meets the agreed bar, we staff up and deliver in regular batches with ongoing QA.
Frequently asked questions
How do you make sure collected data matches our requirements?
Every program starts with a written specification covering format, metadata, contributor profile and acceptance criteria. We run a small pilot, review the output with you, and only scale once the pilot meets the agreed bar.
Can you collect data in specific languages or regions?
Yes. Contributors are recruited to match the languages, locales and demographics the project needs. See our multilingual data service for more on language coverage.
Who owns the collected data?
Data ownership and usage rights are defined in the project agreement before collection begins. Projects can run under NDA.
Related services
- Speech & Voice DataConversational and scripted speech, accents, emotion and professional voice talent.
- Multilingual AI DataNative-speaker collection, transcription, localization and evaluation across languages.
- Data Annotation & LabelingImage, video, text and audio labeling with written guidelines and layered quality review.
Planning a data collection project?
Tell us what your model needs. We'll come back with questions, a proposed approach and a pilot plan.