ConversationalSinhalaCustom-collected
Sinhala Speech Dataset
Short answer: Looking for a Sinhala speech dataset? Spirelight records one to your specification, so you set the speaker mix, dialect, scenario, and recording conditions. Feasibility, sample status, target volume, rights, schedule, acceptance criteria, and final price are confirmed against your brief, and a free sample comes back within 48 hours.
Sinhala speech dataset. Recorded to your specification. Speaker profile, formats, rights, schedule, and price are confirmed in the written proposal.
Overview
Sinhala Speech Dataset is recorded to your specification: we run the Sinhala speech and video collection against your brief rather than reselling an off-the-shelf corpus, so speaker mix, dialect, scenario, and recording conditions are yours to set.
It can be scoped for Speech recognition, Voice agents, LLM training when those requirements are confirmed in the project brief.
The written proposal confirms recruitment feasibility, target volume, recording conditions, annotation, QA, usage rights, delivery artifacts, schedule, acceptance criteria, and final price.
Specification
- Language
- Sinhala
- Target location
- Sri Lanka
- Planned formats
- 16bit 48kHz wavmkvjson
- Planned delivery
- Cloud storageDirect downloadSFTP
- Planned frequency
- On requestweeklymonthly
- Target use cases
- Speech recognitionVoice agentsLLM training
- Categories
- Conversational
- Intended buyers
- Small businessMedium-sized businessEnterprise
Indicative pricing
Tell us the language, volume, recording conditions, rights, and schedule you need and we will send a written quote. Availability, scope, and timeline are confirmed per brief.
Free sample or written quote
Free sample
Tell us what you are building and we will email you samples from this dataset within 48 hours, selected by hand to match the conditions you need to hear. No account, no verification code, no obligation.
Price quote
Tell us about the target volume, speakers, conditions, annotation, rights, and schedule. Verify your work email below to submit the request for feasibility review and a written proposal.
Emotion and affect labeling
Emotion labeling can be scoped as an annotation layer for this configuration, delivered with per-class inter-annotator agreement once confirmed in the project brief.
- Discrete classes can be tagged per utterance or speaker turn: anger, disgust, fear, happiness, sadness, surprise, and neutral.
- Valence and arousal ratings can be layered on the same segments when continuous targets suit the objective better.
- Intent can be kept separate, delivered as dialogue acts or mapped to a taxonomy you define.
- This configuration is specified as spontaneous conversation rather than read prompts, so the recordings can be labeled for affect that occurs rather than affect that is performed.
- Where synchronized video is part of the planned delivery, affect can be labeled from face and voice together, not audio alone.
Frequently asked questions
What is Sinhala Speech Dataset?
Sinhala speech dataset. Recorded to your specification. Speaker profile, formats, rights, schedule, and price are confirmed in the written proposal.
What can Sinhala Speech Dataset be scoped for?
Sinhala Speech Dataset can be scoped for Speech recognition, Voice agents, LLM training when those requirements are confirmed in the project brief.
Which target locations are listed for Sinhala Speech Dataset?
Sinhala Speech Dataset is recorded in Sri Lanka. This does not assert current speaker availability; recruitment feasibility is confirmed for the brief.
How is delivery scoped for Sinhala Speech Dataset?
Sinhala Speech Dataset is delivered over Cloud storage, Direct download, SFTP. Planned formats: 16bit 48kHz wav, mkv, json. Final formats, fields, transfer method, and delivery schedule are confirmed in the written proposal.
Other collection configurations
Guides for buyers
Not sure this is the right configuration? Send the brief and we will confirm feasibility for the language, dialect, and recording conditions you need, or point you to a configuration that fits.