AI · Collection · Hebrish
Parakeet V2 Fine-Tune - Training Data
Audio the Parakeet v2 fine-tunes blend from: master dataset, category and task subsets, imports, one blend dataset per model.
28 datasets
- dataset danielrosehill/Read-Aloud-Voice-Sentences Master dataset: every take from the capture app. Subsets below are copies of its rows.
- dataset danielrosehill/English-Hebrew-Codeswitching-Sentences Subset `code_switch`: English sentences with everyday Hebrew words and Israeli names written in Latin letters, read aloud by one English speaker living in Israel.
- dataset danielrosehill/Tech-Sentences Subset `tech`: 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.
- dataset danielrosehill/AI-Interactions-Sentences Subset `ai_interactions`: 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, rebases.
- dataset danielrosehill/Linux-Admin-Sentences Subset `linux`: Dictation-style English sentences about Linux desktop and server administration (KDE Plasma, GNOME, systemd, storage, networking, homelab), read aloud by one speaker.
- dataset danielrosehill/Smart-Home-Sentences Subset `smart_home`: Dictation-style English sentences about home automation (Home Assistant, Zigbee, Z-Wave, Matter, Thread, ESPHome, MQTT and device brands), read aloud by one speaker.
- dataset danielrosehill/Personal-Vocabulary-Sentences Subset `personal`: 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 Herman Poppleberry), read aloud by that speaker.
- dataset danielrosehill/Homophone-Context-Sentences Subset `homophone`: 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.
- dataset Hebrish/Yiddish-Sentences Subset `yiddish`: 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.
- dataset Hebrish/Jewish-Religious-Terminology-Sentences Subset `religious`: 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.
- dataset danielrosehill/General-English-Sentences Subset `general`: 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.
- dataset Hebrish/Noisy-Condition-Sentences Subset `noisy`: 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.
- dataset Hebrish/Disfluent-Read-Sentences Task subset `disfluency`: Experimental.
- dataset Hebrish/Self-Correction-Sentences Task subset `self-corrections`: One speaker correcting himself mid-sentence ("send it Tuesday, no, Wednesday").
- dataset Hebrish/Spoken-Aside-Sentences Task subset `asides`: One speaker dictating a sentence and, partway through, saying something to someone in the room ("come in, Hannah").
- dataset Hebrish/Hard-ASR-Sentences Task subset `hard-asr`: 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.
- dataset Hebrish/Brand-Name-Sentences Task subset `brand-names`: 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.
- dataset Hebrish/Acronym-Sentences Task subset `acronyms`: 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.
- dataset Hebrish/Irish-English-Sentences Task subset `irish-english`: 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, Gardaí, sláinte) and Irish names and places (Siobhán, Tadhg, Dún Laoghaire), in standard Irish spelling with fadas.
- dataset Hebrish/Repeated-Word-Sentences Task subset `repetitions`: Experimental.
- dataset Hebrish/Singing-Sentences Task subset `singing`: Experimental.
- dataset Hebrish/Weird-Voice-Sentences Task subset `weird-voices`: Experimental.
- dataset Hebrish/Currency-Sentences Task subset `currency`: One speaker saying amounts of money the natural way ("50 shekels", "a dollar fifty", "6 pounds 50").
- dataset Hebrish/Expanded-Acronym-Sentences Task subset `expanded-acronyms`: 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").
- dataset Hebrish/English-Hebrew-Mixed-Sentences Import: the earlier code-switching recordings, in the master dataset as `ehms_*` rows.
- dataset Hebrish/Parakeet-V2-Blend-All Blend dataset for run `r7-all` (kitchen sink: English-Hebrew code-switching plus tech vocabulary, accent and general accuracy): the exact training mix of Hebrish/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching-All-Subsets.
- dataset Hebrish/Parakeet-V2-Blend-No-Hebrew Blend dataset for run `r7-general` (no Hebrew: tech vocabulary, accent and general accuracy): the exact training mix of Hebrish/Parakeet-TDT-0.6B-v2-Tech-And-Accent.
- dataset Hebrish/October-7-Memorial-Sentence-Bank
Copied from Hugging Face on 7 November 2026.