Recognize natural speech
Specify regional language coverage, accents, code-switching and acoustic conditions. Assess transcript conventions against your recognition task.
FIUND / Speech & voice AI
Speech data that reflects how people actually talk.
License conversations and language-specific recordings for speech recognition, speaker understanding and conversational AI. Define the speaker mix, recording conditions and evaluation task before selecting the data.
FIUND / COLLECTION STRUCTURE
01
Recording
sourceConversation context
02
Speaker turns
segmentsBoundaries · roles · overlap
03
Written representation
transcriptLanguage · timestamps · text
ILLUSTRATIVE FORMAT Assets confirmed per collection
Model tasks
Choose data for the behavior you need.
Understand who spoke when
Look for useful speaker separation, interruptions and overlapping conversation. Agree the timestamp and speaker-label conventions needed for your evaluation.
Evaluate real conversations
Use connected exchanges to examine turn-taking and spoken context. Define the setting and participants that matter for your model.
Relevant data types
Explore the source formats and context that could support your task. Access and suitability are confirmed for each proposed project.
Scope before scale
Define a sample you can evaluate.
Language and setting
Countries, varieties, speaker characteristics, recording environments and channel conditions.
Media and annotations
Source audio, track separation, sampling rate, transcript availability and any additional annotation work.
Evaluation and permissions
Representative samples, intended use, acceptance criteria and permitted recipients.
Sample review
Questions to settle early
- Language coverage matches the brief
- Speaker and transcript references are interpretable
- Recording conditions are disclosed
- Sample permissions are explicit
Start with a collection.
Review the published collection scope, then request a sample for your intended evaluation. Other sources and annotations are scoped separately.
Frequently asked questions
Are transcripts and speaker labels always included?
They depend on the selected collection. We confirm existing annotations and identify any preparation required before agreeing scope and pricing.
Can I request speech in another language?
Yes. Specify the region, recording setting and speaker mix. Suitable sources, permissions and timing are confirmed against your brief.
Does a speech licence permit voice cloning?
The standard licence does not authorize cloning an identifiable person’s voice or likeness. Permitted AI uses are defined in the agreement.