SilencioNetwork/tagalog-filipino-speech

Tagalog / Filipino Spontaneous Speech — Silencio Philippines Pack

Spontaneous Tagalog/Filipino with human transcription and word-level forced alignment. Sixteen speakers, 90 unscripted clips, 13,782 timestamped tokens. Part of the Silencio Philippines Pack.

Hours	2.03
Clips	90
Speakers	16
Countries	1
Speaker origin regions	2
L1 speakers of the recorded language	15 of 16 (86 clips)
Audio	48 kHz stereo WAV
Mean clip length	81.0 s
Transcripts	human_validated: 90
Licence	cc-by-nc-4.0

All 90 clips carry a human transcription.

Recordings are unscripted responses to open prompts, captured on contributors' own devices in their own environments. Mean clip length is 81 seconds — spontaneous speech, not read utterances.

Load it

from datasets import load_dataset

ds = load_dataset("SilencioNetwork/tagalog-filipino-speech", split="train")
print(ds[0]["transcript"], ds[0]["dialect"], ds[0]["country"])

# datasets v4 returns a torchcodec AudioDecoder:
s = ds[0]["audio"].get_all_samples()
audio, sr = s.data, s.sample_rate

Requires pip install "datasets>=4.0" and FFmpeg ≥ 4.

Speaker and recording metadata

By country

Country	Speakers	%
Philippines	16	100.0%

Speaker origin / self-reported variety — this is the speaker's own background, not a dialect classification of the recorded language

Speaker origin	Speakers	%
Philippines - Manila (Tagalog)	14	87.5%
Philippines - Filipino English (Manila)	2	12.5%

Demographics

Gender	Speakers	%
male	15	93.8%
female	1	6.2%

Age band	Speakers	%
45-59	6	37.5%
25-34	5	31.2%
35-44	4	25.0%
18-24	1	6.2%

Recording conditions

Device	Clips	%
Mobile	81	90.0%
Desktop	9	10.0%

Splits

Single split, test, 90 rows. No train/dev/test partition is provided: at this scale a partition would leave each part too small to be meaningful. Speaker identifiers are stable, so a speaker-disjoint split can be constructed at load time.

Fields

Column	Description	Values in this release
audio	Audio payload. Stored at source rate; see the spec table for the exact distribution	48 kHz stereo WAV
speaker_id	Pseudonymous speaker identifier. Coherent within this dataset; deliberately not linkable to other Silencio releases	16 distinct
language	Language of the recording	constant: Tagalog / Filipino
transcript	Human transcription of the recording	90 distinct
transcript_type	Provenance of the transcript	constant: human_validated
gender	Self-reported	female, male
country	Speaker's country	constant: Philippines
mother_tongue	Speaker's self-reported first language	English, Tagalog / Filipino
dialect	Self-reported speaker origin / regional variety. This is the speaker's own background, NOT a dialect classification of the recorded language	Philippines - Filipino English (Manila), Philippines - Manila (Tagalog)
os	Operating system of the recording device	Linux, macOS
device	Recording device class	Desktop, Mobile
duration	Seconds	90 distinct
script_type	Elicitation style	constant: free_speech
words	Word-level forced alignment: text, normalised text, start and end in seconds	13,782 entries across 90 clips
n_words	Number of aligned tokens in this clip	66 distinct
transcript_model	How the transcript text was produced	constant: human
aligner	Model used to produce the word timings	constant: wav2vec2_mms
age_band	Self-reported age, banded	18-24, 25-34, 35-44, 45-59
native_speaker	True where mother_tongue matches the recorded language	2 distinct
proficiency	Speaker's self-declared proficiency in the recorded language	fluent, native

Related Tagalog and Filipino speech resources

Tagalog, in its standardised form Filipino, is the national language of the Philippines and the first language of roughly 30 million people. Unlike Cebuano or Hiligaynon it already has meaningful Hub coverage, so this release is positioned against it rather than as a first:

Resource	Scale	Type	Licence
google/fleurs (fil_ph)	3,266 utterances	Read Wikipedia sentences	CC BY 4.0
disco-eth/WorldSpeech	79 languages	Mixed-source multilingual	—
sapinsapin/filipinospeechcorpus	6.9 GB	Filipino speech corpus	MIT
RidheshBhati/filipino-tts-biblical	Small	Scripture, TTS-oriented	—
This dataset	90 clips, 2.0 h, 81 s mean	Spontaneous, word-level alignment, speaker metadata	CC BY-NC 4.0

What this adds to an already-served language is unscripted speech with per-word timings and speaker-level metadata. FLEURS and the biblical corpora are read speech; models tuned on them degrade on spontaneous input, and that gap is what this sample measures.

Also from Silencio. Cebuano is published under this same protocol. Hiligaynon and expanded Tagalog follow — see SilencioNetwork.

Transcription and alignment

Two distinct provenances, kept separate because they carry different confidence.

Text — human. Every transcript was produced by a human annotator listening to the recording, recorded per clip in transcript_model.

Timings — machine. Word-level start and end times come from forced alignment with wav2vec2_mms, recorded per clip in aligner. On every clip the aligner's token count matches the human reference token count exactly, and no word timing runs past the end of its audio file.

The words column holds one entry per token with text, normalized_text, start and end in seconds:

ds = load_dataset("SilencioNetwork/tagalog-filipino-speech", split="test")
for w in ds[0]["words"][:5]:
    print(f"{w['start']:6.2f}-{w['end']:6.2f}  {w['text']}")

Speaker proficiency

Proficiency is taken from each contributor's own declared language profile rather than inferred from a single primary-language field.

Declared Tagalog/Filipino level	Speakers	Clips	Hours
native	15	86	1.98
fluent	1	4	0.04

Filter on native_speaker, or on proficiency for finer control.

What this is useful for

• Spontaneous-speech ASR evaluation. Unscripted Tagalog with human reference text, in a language whose public data is dominated by read speech.
• Forced-alignment and VAD work. 13,782 word-level timings across 2 hours.
• Code-switching. Spontaneous Manila Tagalog contains heavy English admixture — Taglish is the everyday register. Not annotated as such in this release, but present throughout.
• Contact-centre and voice-agent robustness, where Philippine-accented spontaneous speech is the operating condition rather than the edge case.

Limitations

• Severe gender imbalance. 15 of 16 speakers are male. This is the most significant limitation of the sample and it will bias anything trained or evaluated on it. Gender-balanced Tagalog is available in the wider inventory and through the collection programme; it is simply not in this sample. Do not use this release for any gender-comparative work.
• Sample scale. 90 clips, 16 speakers, 2.0 hours. Suitable for evaluation and alignment work; not a training corpus.
• Speaker origin is concentrated. All contributors record a Metro Manila origin. Regional Tagalog varieties — Batangas, Bulacan, Quezon, Rizal — are not represented in this sample.
• Proficiency is self-declared and not independently assessed.
• Word timings are machine-generated. Forced alignment with wav2vec2_mms, not manually corrected. Token counts reconcile exactly against the human reference on every clip, but individual boundaries have not been human-verified.
• Code-switching is unmarked. English admixture is pervasive and is transcribed as spoken, with no language tagging at token level.
• No acoustic annotation. Recording environment, background-noise class and SNR are not annotated. Available for commissioned collection.
• Mixed audio format. Source audio is shipped untouched at its captured sample rate and channel count — see the spec table. Resample and downmix before batching.
• No diarisation. Single speaker per clip. Multi-speaker material is part of the collection programme described below, not this release.
• Pseudonymous speakers. speaker_id values are pseudonyms, coherent within this dataset, deliberately not linkable to speakers in other Silencio releases.
• No baseline. No reference WER is published with this release.

Provenance and consent

Every recording is contributed by an opted-in participant through the Silencio app, under a consent record covering AI/ML training use. Contributors can request deletion, and deletion propagates to downstream releases. Full provenance documentation is available to licensees.

License

cc-by-nc-4.0 — free for research and non-commercial use with attribution.

Attribution string: Silencio Network, Tagalog / Filipino Spontaneous Speech, 2026. CC BY-NC 4.0.

Non-commercial covers research, evaluation and publication. Benchmarking a commercial product model against this data is a commercial use and needs a licence — ask, it is usually granted for evaluation. Model weights trained on this sample inherit the non-commercial restriction. Contributors may withdraw consent; withdrawal propagates to subsequent releases but places no retroactive obligation on an existing licensee.

Commercial licensing, including terms for models trained on this data: info@silencio.network

Citation

@misc{silencio_tagalog_2026,
  title  = {Tagalog / Filipino Spontaneous Speech — Silencio Philippines Pack},
  author = {Silencio Network},
  year   = {2026},
  url    = {https://huggingface.co/datasets/SilencioNetwork/tagalog-filipino-speech}
}

The Silencio Philippines Pack

Philippine inventory as of August 2026:

Language / variety	Hours	Recordings	Speakers
Tagalog / Filipino	2,135	192,264	5,507
Philippine English	2,705	129,433	2,340
Cebuano	505	31,668	614
Hiligaynon	12	1,600	51
Ilocano	11	1,186	50

In active collection: 7,500 hours. A collection and human-transcription programme covering 2,500 hours each of Tagalog, Cebuano and Hiligaynon, split per language into 1,000 hours single-speaker and 1,500 hours multi-speaker.

Language	Single-speaker	Multi-speaker	Total
Tagalog	1,000 h	1,500 h	2,500 h
Cebuano	1,000 h	1,500 h	2,500 h
Hiligaynon	1,000 h	1,500 h	2,500 h
Total	3,000 h	4,500 h	7,500 h

Ilocano, Waray, Bikol, Kapampangan and Pangasinan are available through commissioned collection.

Silencio corpus and collection network

Two distinct figures, because they answer different questions.

Recorded and available off the shelf — audio already collected, with metadata, licensable today:

Hours recorded	127,793
Recordings	9,392,870
Contributors who recorded	222,145
Languages	156
Countries and territories of origin	216

Contributor network available for commissioned collection — registered, consented contributors who can be activated for a specific brief. These are not active contributors to the corpus above; they are the pool it is drawn from and extended through:

Registered contributors	2,000,000+
Countries	180+
Languages reachable	250+

For volume licensing, pre-release access to the Philippines programme, or commissioned collection in a language not listed: info@silencio.network