AI · datasets · audio
Audio datasets
Every audio dataset I've published, in one list, whichever Hugging Face account it sits on. These are recordings of my own voice for speech recognition and text-to-speech, unless a set says otherwise.
- 39
- datasets in the collection
- 16 + 23
- personal account + Hebrish
- 40.3 h
- in the master dataset
- ~44 h
- unique audio in total
How the sets fit together
Takes go into one master dataset, Read-Aloud Voice Sentences. Most of the others are slices of it by topic or condition, so a take can appear in several and their hours overlap: don't add them up. Sets marked Standalone hold recordings that aren't in the master. Gated sets need an access request on Hugging Face.
Master (1)
- Read Aloud Voice Sentences Read-aloud sentences from one speaker, recorded on a phone (OnePlus Nord 3 5G, internal microphone).
Everyday and general (6)
- General English Sentences Plain, everyday English sentences (emails, messages, general prose and homophone pairs) read aloud by one speaker, with no Hebrew or Yiddish words, Israeli names or tech vocabulary. A small, general set for adapting a model to…
- Irish English Sentences Everyday sentences in Irish English read by one speaker born in Ireland: vocabulary (craic, grand, giving out, press, messages), grammar (I'm after doing it, so it is, ye), Irish-language words used in English (Taoiseach, Dáil…
- Personal Vocabulary Sentences English sentences built around names a speech model gets wrong for one speaker: his own name, his company and his project names (Daniel Rosehill, DSR Holdings, Carrot Cake AI, the podcast My Weird Prompts and its hosts Corn and…
- Voice Command Sentences One speaker reading desktop voice commands for a KDE Plasma voice-command agent: launching, switching to and closing apps (Konsole, Kate, Dolphin, Kdenlive, Audacity), browser and terminal tabs, windows and virtual desktops…
- Smart Home Sentences Dictation-style English sentences about home automation (Home Assistant, Zigbee, Z-Wave, Matter, Thread, ESPHome, MQTT and device brands), read aloud by one speaker. Includes product names a speech model writes as an ordinary…
- AI Interactions Sentences Instructions dictated to AI coding agents (Claude Code, Codex and others) and the git and GitHub workflow around them, read aloud by one speaker: subagents, worktrees, compaction, plan mode, hooks, MCP servers, pull requests…
Tech (6)
- Tech Sentences Dictation-style English sentences dense in software, DevOps, cloud, self-hosting and AI vocabulary (product names, CLI tools, acronyms, model names), read aloud by one speaker. Terms use the vendor's own casing (GitHub…
- Tech Sentences For ASR Training A specialized speech dataset for fine-tuning Automatic Speech Recognition (ASR) models on technical and developer vocabulary. Contains human-recorded audio samples covering software engineering, cloud infrastructure, DevOps, and…
- Linux Admin Sentences Dictation-style English sentences about Linux desktop and server administration (KDE Plasma, GNOME, systemd, storage, networking, homelab), read aloud by one speaker. The focus is names that a speech model writes as an ordinary…
- Acronym Sentences Natural sentences with acronyms read by one speaker, both spelled out (API, SSH, HTTPS) and said as words (NASA, JSON, YAML), each in its standard written form. prompt_kind is letters, word or mixed.
- Expanded Acronym Sentences Chat and business shorthand said as letters by one speaker ("by E O D", "F Y I", "I M O") with the transcript writing it out in full ("by end of day", "For your information", "In my opinion"). spoken_text is the script with the…
- Brand Name Sentences Natural sentences with company, product, AI model and open-source project names (Anthropic, OpenRouter, Amazon Bedrock, Qwen, Kubernetes, Nginx) read by one speaker, written in each name's official form. target_terms lists the…
Hebrew, Yiddish and Jewish life (7)
- English Hebrew Codeswitching Sentences English sentences with everyday Hebrew words and Israeli names written in Latin letters, read aloud by one English speaker living in Israel. Hebrew words follow one canonical spelling convention, and Hebrew proper nouns are…
- English Hebrew Mixed Sentences A dataset of English sentences with Hebrew words and phrases interspersed, designed for speech-to-text training and evaluation for English speakers in Israel.
- Hebrish TTS Voice Recordings Single-speaker read and spontaneous speech for training a text-to-speech voice that code-switches between English and Hebrew ("Hebrish") the way an English-speaking immigrant in Israel does.
- Yiddish Sentences English sentences with Yiddish words written in Latin letters (shul, daven, cholent, mensch, freier, tachles), the way English speakers in Israel use them, read aloud by one speaker. Many of these words are also used in modern…
- Jewish Religious Terminology Sentences English sentences with Jewish religious vocabulary written in Latin letters: Shabbat and the festivals, prayer and the synagogue, Torah portions, Talmud and halacha, kashrut, and life-cycle events, read aloud by one speaker. Many…
- Judaica Sentences Real sentences from traditional Jewish works in English, read by one speaker, Hebrew terms in romanized form: Tanach (Metsudah), Mishnah Yomit (Joshua Kulp), Talmud Bavli (Sefaria Community Translation; Tractate Berakot, A.…
- Israeli Name Sentences English messages and emails that mention an Israeli or Jewish person's name, a first name (Naftali, Yotam, Tzlil, Bat-El) or a surname (Buskila, Ifergan, Weinberg), written in its common Israeli-English spelling, read aloud by…
Hard words and numbers (4)
- Hard ASR Sentences English sentences speech recognition finds hard, read by one speaker: Harvard sentences (IEEE 1969, public domain), tongue twisters, homographs, garden-path sentences, rare words, dense numbers and units, unusual proper nouns.…
- Homophone Context Sentences English sentences whose context settles which spelling of a homophone is meant (their/there, principal/principle, and 58 more sets), read aloud by one speaker. Sentences to read: Homophone-Context-Sentence-Bank.
- Pronunciation Variants Takes in which a word was deliberately said one particular way, from the main dataset, stratified by variant_category. Two kinds: a term from the project's pronunciation registry said in a recorded variant…
- Currency Sentences One speaker saying amounts of money the natural way ("50 shekels", "a dollar fifty", "6 pounds 50"). spoken_text is the script that was read; text is the written form a dictation model should give: 50 NIS, $1.50, €18, £6.50 (NIS…
Speech conditions (8)
- Noisy Condition Sentences Every take from the main dataset recorded in noise, whatever its category: outdoors, in a noisy room, with traffic, other people talking, background noise or a crying baby. A robustness and evaluation set for dictation in real…
- Whispered Sentences Experimental robustness data. Plain English sentences, short and dictation-length, whispered by one speaker into a phone, the way he dictates when someone is asleep nearby.
- Disfluent Read Sentences Experimental. Plain English sentences read with deliberate hesitations (uh, um, eh; repeated words have their own dataset since 2026-10-20) by one speaker.
- Repeated Word Sentences Experimental. Plain English sentences read by one speaker repeating a word or phrase now and then ("the the report", "I think I think"), with no fillers.
- Self Correction Sentences One speaker correcting himself mid-sentence ("send it Tuesday, no, Wednesday"). spoken_text is the script that was read; text is the corrected sentence only, the transcript a dictation model should give.
- Spoken Aside Sentences One speaker dictating a sentence and, partway through, saying something to someone in the room ("come in, Hannah"). text excludes the aside; aside_text and aside_position (start, middle, end) give it; spoken_text is the full…
- Singing Sentences Experimental. Short plain English sentences sung by one speaker, to any tune.
- Weird Voice Sentences Experimental. Plain English sentences read by one speaker in a put-on voice (very deep, squeaky, robot, cartoon, mock accent, mumbly).
Evaluation and task sets (5)
- English Hebrew Codeswitching ASR Benchmark A frozen, versioned test set for measuring how well speech recognition handles English speech with Hebrew words in it: "pick up some bamba from the makolet", "I have to go to Bituach Leumi tomorrow". The references write each…
- Hebrish Validation A validation benchmark for the Hebrish English-Hebrew code-switching speech recognition models (Sabich and the other lines). Each row is one continuous read of a fixed script of about one minute, with the script as the exact…
- Small STT Eval Audio Dataset A small speech-to-text evaluation dataset containing 92 audio samples with ground truth transcriptions. Designed for evaluating STT systems on technical vocabulary, code-switching (English/Hebrew), and various speaking styles.
- Paragraph Dictation Passages Data for a separate paragraph-inference model; not used to train any Hebrish model. Passages of 2 to 4 paragraphs (emails, prompts to AI coding agents, voice notes, messages, meeting notes, blog drafts) dictated in one take by…
- Proofed ASR Cleanup Training Data Audio paired with clean, proofed target text for training an ASR model to do its own post-processing: no fillers, repeats or false starts, numbers and money as digits, self-corrections applied, asides to other people left out…
Voice notes and narration (2)
- Annotated Voice Notes Hand-picked real voice notes, each with audio, a raw AssemblyAI transcript and a note on why it was included.
- Sample Voice Context Data A job seeker’s career story, written by an LLM and narrated, made to test a voice-to-vector-database RAG pipeline.
Earlier sets and experiments (10)
Older evaluation sets and one-off experiments, about 6 hours more. They're outside the Hugging Face collection and the total above, and I haven't checked them for overlap with each other.
- Voice Note Audio Real-world voice notes with transcripts and annotations.
- STT Voice Notes Evals Voice notes for comparing speech-to-text services, from August 2025.
- ASR WPM And Background Noise Eval Annotated clips for testing how speaking rate and background noise affect transcription accuracy.
- Audio Understanding Test Set A voice sample and 137 test prompts for evaluating audio understanding in multimodal models.
- Long Prompt Experiment One long spoken prompt, from an experiment on prompt structure and optimisation.
- Noisy Voice Notes Real voice notes with background noise: traffic and everyday places, recorded over about a year.
- Audio Understanding Bitrate Eval 0426 How MP3 bitrate affects transcription accuracy across audio-input LLMs, April 2026.
- Wake Words Jewish and Israeli first-name wake words for local wake word models. Several speakers, not only me.
- Transcription Cleanup Trainer A recording with raw and manually cleaned transcripts, for training transcript cleanup.
- Whisper Fine Tune One Shot Eval A quick comparison of fine-tuned local Whisper models against commercial ASR APIs.
Hours and takes as of 7 November 2026, counted from the repositories' own files. Totals are rounded. ← All datasets