FalAR-TTS: European Portuguese Speech Recordings from 2016-2017

FalAR-TTS is a subset of the FalAR dataset tailored for speech synthesis in European Portuguese. The subset is restricted to utterances recorded between 2016 and 2017 to minimize longitudinal acoustic variations. It was created by inesc-id and features additional filtering for highly reliable timestamped speech alignments.

Use Cases

• Train text-to-speech models based on European Portuguese audio recordings.
• Develop speech synthesis systems using timestamped speech alignments.
• Research acoustic variations in speech signals over time within the 2016-2017 period.

Strengths

• Recordings are from a controlled time range (2016-2017) to minimize acoustic variation.
• Filtering criteria were applied to ensure highly reliable timestamped speech alignments.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Subset of the FalAR dataset (https://huggingface.co/datasets/inesc-id/FalAR) by inesc-id.

Collection Method

Likely contains speech recordings, but specific collection method is not detailed.

Time Range

Recordings are from between 2016 and 2017.

Freshness

Last updated 2026-06-29 15:12:08; freshness should be verified.

Geography

Data is in European Portuguese, suggesting a primary focus on Portugal.