Luel Conversational Speech Samples in Three Quality Tiers

Multi-speaker conversational audio for ASR, TTS, and dialogue research, organized into three quality tiers. The dataset was created by Luel-ai and was last updated on 2026-05-20. It includes a curated standard tier in 9 languages and a noisy tier for English.

Use Cases

• Training automatic speech recognition (ASR) models based on multi-speaker conversational audio.
• Developing text-to-speech (TTS) systems based on curated speech samples.
• Researching dialogue systems based on conversational audio data.
• Building noise-robust ASR models based on the 'noisy' tier recorded in non-ideal acoustic environments.

Strengths

• Organized into three distinct quality tiers, including a curated 'Standard' tier.
• The 'Standard' tier covers 9 well-resourced languages.
• Includes a dedicated 'Noisy' tier for training noise-robust models.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• The 'Noisy' tier is currently English-only, limiting its application for other languages.

Provenance

Source

huggingface, author Luel-ai

Collection Method

Likely contains recorded conversational audio, with some tiers undergoing curation and QA.

Time Range

null

Freshness

Last updated 2026-05-20 06:44:32; freshness should be verified.

Geography

null