65 hours of my own voice: an index of my speech datasets

64.7 hours and 46,227 takes of my own voice as of 13 November 2026: how the audio datasets on Hugging Face fit together, the hours in each, and the models they feed (Sabich, Fritter, Falafel).

Daniel Rosehill

(3 Kislev 5787)

24 min read

65 hours of my own voice: an index of my speech datasets

A few weeks ago, recording my own voice for speech models was something I did for an hour here and there. Then it became a daily habit, and at some point I lost track of how much there actually is. So I counted.

As of today, 13 November 2026, the master dataset holds 64.7 hours of my voice across 46,227 takes. That's the number I'd been loosely calling "about 60 hours", and it turns out I was underselling it. (The growth chart below is a fair picture of what the last month has looked like.)

Cumulative hours of recorded audio in the master dataset, 13 October to 13 November 2026, reaching 64.7 hours, with hours recorded per day below

How the data is organised

Everything starts in one place. Every take I record in the capture app lands in a single master dataset, Read-Aloud-Voice-Sentences on Hugging Face, and every speech model I train draws from it.

Hugging Face Dataset danielrosehill/Read-Aloud-Voice-Sentences ↓ 40.8kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face

The other datasets are slices of that master, not separate recordings. Tech-Sentences is the takes dense in software and DevOps vocabulary, English-Hebrew-Codeswitching-Sentences is the ones with Hebrew words mixed in, and so on. Some slices have slices of their own (Linux-Admin-Sentences sits inside Tech-Sentences, for instance). Because a take can belong to several subsets at once (a noisy take about Kubernetes is both "noisy" and "tech"), the hours below overlap and don't add up to the total. The 64.7 hours in the master is the real figure.

The biggest slices right now:

Dataset                                  Hours   Takes
Read-Aloud-Voice-Sentences                64.7  46,227
Tech-Sentences                            16.9  10,944
English-Hebrew-Codeswitching-Sentences    15.8   9,489
Noisy-Condition-Sentences                 10.1   7,202
Voice-Command-Sentences                    5.5   7,778
General-English-Sentences                  4.3   3,013
Judaica-Sentences                          2.0   1,034
Israeli-Street-Sentences                   1.6   1,052

Noisy-Condition-Sentences is the odd one out: it groups takes by how they were recorded (in noise, whatever the topic) rather than by what they say.

What it feeds

The same recordings sit underneath every model line:

  • Sabich, the main Hebrish model, trains on the full mix and is updated as a rolling checkpoint.

  • Jachnun and Shakshuka (the streaming model) are the code-switching line.

  • Fritter, my English dictation model for technical vocabulary, trains on the takes without code-switching.

  • Falafel, the small voice-command model, trains mainly on Voice-Command-Sentences.

  • Meorav and Fritter-TTS, the text-to-speech voices, use the takes clean enough for speech synthesis.

So when I record a hundred sentences about API work on a Tuesday evening, they end up in several models at once. That's most of the reason I keep everything in one master rather than in separate per-project datasets.

Every audio dataset, by hours

Here is the whole index, largest first. Hours and take counts are as of 13 November 2026, and they will be out of date within a day or two, because I record most days and the datasets sync automatically. For the current numbers, the Audio-Dataset-Index on Hugging Face is rebuilt from the same source.

Hugging Face Dataset danielrosehill/Audio-Dataset-Index Every audio dataset Daniel Rosehill publishes on Hugging Face, on this profile and in the Hebrish organisation (the name of the English-Hebrew speech project), as one tree: which dataset is a superset of which, which… ↓ 58CC-BY-4.0Updated Nov 2026 View dataset on Hugging Face

Most are public. The Hebrew ones, and a few others, are gated (access by request), because there's some potential for misuse with Hebrew speech data specifically.

Hugging Face Dataset danielrosehill/Tech-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Superset of… ↓ 3k♥ 1Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/English-Hebrew-Codeswitching-Sentences ↓ 38.7k♥ 1Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Noisy-Condition-Sentences ↓ 36.6kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Voice-Command-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 665Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/General-English-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Superset of… ↓ 2.7kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Jewish-Religious-Terminology-Sentences ↓ 34.8kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Judaica-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Superset of Hebrish/Tanach-Verse-Sentences:… ↓ 1.3kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Israeli-Street-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 426Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Linux-Admin-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of danielrosehill/Tech-Sentences: its… ↓ 1.1kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/AI-Interactions-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of danielrosehill/Tech-Sentences: its… ↓ 1.4kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Tanach-Verse-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of Hebrish/Judaica-Sentences: its… ↓ 104Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Personal-Vocabulary-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 1.4kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Brand-Name-Sentences ↓ 410Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Claude-Dictations Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 313Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Date-Time-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 324Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Yiddish-Sentences ↓ 19.6kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/AI-Integration-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 394Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Numerics-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 233Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Whispered-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 248Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Open-Source-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 376Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Irish-English-Sentences ↓ 9kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/AI-Model-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 389Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Smart-Home-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of danielrosehill/Tech-Sentences: its… ↓ 1.4kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Email-Address-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 330Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Web-Address-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 264Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/GitHub-Trending-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 373Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Israeli-Name-Sentences ↓ 16.3kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Dev-Stack-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 347Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Pronunciation-Variants ↓ 7.2kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Currency-Sentences ↓ 5.2kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Homophone-Context-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of… ↓ 612Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Israeli-Place-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 338Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Disfluent-Read-Sentences ↓ 460Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Phone-Number-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 189Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Israeli-Business-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 185Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Idiom-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 226Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Procurement-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 187Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Entity-Dense-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 121Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Colloquial-Reduction-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 89Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Data-Tool-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 265Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Acronym-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 632Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Shingle-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 257Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/API-Ops-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 236Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/NATO-Alphabet-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 133Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Architecture-Practice-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 126Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Long-Tech-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 68Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Functional-GI-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 198Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Hard-ASR-Sentences ↓ 2.2kAutomatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Latin-Phrase-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 143Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Israeli-TV-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 160Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Hardware-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 143Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Expanded-Acronym-Sentences ↓ 811Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Spelled-Out-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 238Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Repeated-Word-Sentences ↓ 684Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Fragment-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 147Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Long-Code-Switching-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 63Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Self-Correction-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 229Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Fast-Speech-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 67Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Singing-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 288Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Background-Speech Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Subset of Hebrish/Noisy-Condition-Sentences:… ↓ 128Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Thinking-Pause-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 67Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Spoken-Aside-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 221Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Paragraph-Dictation-Passages Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 68Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Weird-Voice-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 229Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/False-Start-Sentences Audio dataset, split-out. Rows copied from the master, danielrosehill/Read-Aloud-Voice-Sentences, so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours:… ↓ 70Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face

Outside the master

A few audio datasets aren't slices of the master: earlier recordings from before the capture app existed, contributions from other people, and experiments.

Hugging Face Dataset Hebrish/Hebrish-TTS-Voice-Recordings ↓ 34.9kText to speechCC-BY-NC-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/English-Hebrew-Mixed-Sentences ↓ 187MITUpdated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Community-Voice-Contributions Audio dataset, standalone. Kept apart from the master dataset and its split-outs. Every audio dataset as one tree, with hours: the audio index. How every dataset fits together: the dataset index. <!--… ↓ 112Automatic speech recognitionCC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset Hebrish/Synthetic-Codeswitching-Voices-Pilot Audio dataset, standalone. Kept apart from the master dataset and its split-outs. Every audio dataset as one tree, with hours: the audio index. How every dataset fits together: the dataset index. <!--… ↓ 92CC-BY-4.0Updated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Tech-Sentences-For-ASR-Training Audio dataset, standalone. Kept apart from the master dataset and its split-outs. Every audio dataset as one tree, with hours: the audio index. How every dataset fits together: the dataset index. <!--… ↓ 529♥ 2Automatic speech recognitionMITUpdated Nov 2026 View dataset on Hugging Face Hugging Face Dataset danielrosehill/Wake-Words Real recordings of Jewish and Israeli first-name wake words, for training local wake word models (openWakeWord, microWakeWord) to use with Home Assistant Assist and on the desktop, in place of "Hey Siri" or "OK Google". ↓ 190Audio classificationCC-BY-4.0Updated Oct 2026 View dataset on Hugging Face

What's next

The obvious gap is that almost all of this is one speaker: me. The models need other voices (different accents, genders and communities) far more than they need another ten hours of mine. If you'd like to contribute recordings, the Hebrish site has a page for it.

In the meantime, I'll keep recording a few hours a week and rebuilding the index.

speech-recognition datasets hugging-face hebrish fine-tuning
Daniel Rosehill

Daniel Rosehill

Agentic AI systems architect in Jerusalem. I design and build multi-agent systems, MCP tooling and speech-to-text pipelines, and write up what I learn along the way.

More about me at danielrosehill.com
About Contact GitHub LinkedIn
Creative Commons License CC BY 4.0