# 3.24.0 · Objects, colors and everyday routines

## 3.27.0 — W03 appointments/café
Four new illustrations, 96 dedicated SVGs and 67 MP3s expand s3/s4. Each now has 16 vocabulary targets and 12 guided checks, translated passages, a recorded glossary, answer explanations and a saved production task. All 6,100 prior media files are unchanged. See `implementation/w03/BATCH-NOTES.md` and `implementation/w03/RELEASE-CHECKS.md` for completed tests and remaining review gates. Earlier sections below are historical.

## 3.26.0 — W02 clothing/travel expansion
Four new illustrations, 94 semantic SVGs and 72 recorded clips support s1/s2. Both now have 16 vocabulary targets, 12 form checks, translated passages, recorded glossaries and saved production tasks. See `implementation/w02/BATCH-NOTES.md` and `implementation/w02/RELEASE-CHECKS.md` for evidence and limits. Earlier sections below are historical.

## 3.25.0 batch update
See `implementation/BATCH-01.md` for the first prerequisite/recognition-support repairs and `implementation/RELEASE-3.25.md` for test evidence. Six new recorded support examples bring the indexed audio library to 4,081 clips; prior media remains unchanged. Historical sections below describe earlier releases.

Two new units add 64 targets in eight sets, thirty form checks, four translated passages / twelve questions, four illustrations, 108 SVG cues and 175 recordings. Fifteen demonstratives and sixty present-person forms have expandable recorded references. Synonym distractors are filtered; reading translation use is tracked as support. All original IDs, palette and shared review behavior are retained.

Current totals: 32 Skills units; 457 vocabulary placements; 342 form checks; 64 passages; 192 meaning questions; 991 possible Skills cards; 909 foundation SVGs; 4,075 indexed recordings / 9,724.14 seconds. The beginner route has 37 stages / 40 resources, with 576 vocabulary placements / 512 distinct listed chunks. Counts are not mastery or complete A1/A2 coverage.

See [OBJECTS-ROUTINES.md](OBJECTS-ROUTINES.md) for correctness sources, the A1/A2 domain-gap review and maintenance. BP04/BP05 remain partially authored. **All sections below are historical release notes unless explicitly stated otherwise.**

---

# 3.23.0 · People and identity expansion

Two supported A1 units now teach family/description and origin/nationality/occupation. Added: 64 vocabulary targets in ten sets, twenty models, 24 forms, four messages / twelve questions, two production tasks, a saved fictional-profile planner, four illustrations, 102 SVG cues and 128 recordings. Current indexed audio: 3,900 clips / 9,351.53 seconds. The existing palette, typography, original numbering and shared scheduler remain.

The 35-step route links 38 resources; original IDs are preserved. Beginner audit: 512 placements / 469 distinct listed chunks. See [PEOPLE-IDENTITY.md](PEOPLE-IDENTITY.md) for sources, scope and maintenance. BP04 joins BP01–BP03 as partially authored, not complete. Broader beginner and higher-level coverage still needs work. Earlier figures below are historical snapshots.

---

# 3.22.0 · Beginner teaching expansion

Four authored A1 units now cover alphabet/spelling, number/price/time patterns, calendar/dates and weather/plans. The 33-step route links 36 units and all 96 Word Lab entries. Stable old lesson and br IDs remain unchanged. Added: 137 vocabulary targets in 17 sets, forty models, 48 form checks, eight passages / 24 questions, four production tasks, eight illustrations, 213 SVG cues and 239 recordings. Total indexed audio: 3,772 clips.

See [BEGINNER-FOUNDATIONS.md](BEGINNER-FOUNDATIONS.md) for the current source-informed scope and limitations. The beginner audit now covers 448 listed placements / 417 distinct chunks; not unique lemmas or proof of A1/A2 completeness. BP01–BP03 are partially authored, not complete. Broader phonological, lexical, dialogue and independent-assessment work remains. Earlier figures and planned-only descriptions below are historical snapshots.

---

# 3.21.0 · no new audio claims

The beginner route reuses existing recorded models/cards for cumulative retrieval. All 311 audited explicit vocabulary placements have recording lookups and existing files; referenced model and form-context lookups pass. The library remains 3,533 indexed recordings. This release adds no new narrator or recordings and does not claim a full human pronunciation audit. See BEGINNER-STRATEGY.md for new audio-production requirements.

---

# 3.20.0 — constructions, discourse and practical emails

Added Skills 21–24: se constructions, action stages, connectors and email transfer. New: 32 vocabulary placements, forty bilingual models, 48 forms with explanations, eight passages / 24 questions, four production tasks, twelve recorded contrasts and a saved five-field email planner with overwrite confirmation. Four new alternative answer strings accepted.

Media: eight illustrations, 108 SVG cues, 138 recordings / 483.41 seconds. Total audio: 3,533 clips / 8,591.21 seconds / 52,910,345 bytes. Skills: 24 units, 192 vocabulary placements, 240 form checks, 48 passages, 144 questions and 576 possible review cards. Original lessons, Word Lab, eleven JW situations, Matrix and progress IDs preserved. See DISCOURSE-EXPANSION.md for sources and limitations; not complete zero-to-C2 coverage or certification. Earlier entries below are historical.

# 3.19.0 — connected messages and hypothetical meaning

Added Skills 17–20: relatives, past-perfect sequencing, reported messages and hypothetical clauses. New: 32 vocabulary placements, 40 bilingual models, 48 form checks, eight passages / 24 questions, four production tasks, nine recorded color-chunk comparisons, 48 explanatory feedback notes and ninety reference slots. Eleven additional accepted relative/-se answer strings are tested.

Media: eight illustrations, 108 SVG cues, 219 indexed recordings / 665.38 seconds. Audio library: 3,395 clips / 8,107.8 seconds / 49,956,394 bytes. Skills: twenty units, 160 vocabulary placements, 192 form checks, forty passages, 120 meaning questions and 472 possible review cards. Word Lab, eleven JW situations, Matrix and original course remain intact.

Route: 1 → 2 → 3 → 4 → 5 → 8 → 9 → 13 → 14 → 6 → 18 → 7 → 10 → 11 → 12 → 15 → 16 → 17 → 19 → 20. See NARRATIVE-EXPANSION.md for sources, teaching boundaries and maintenance. Not complete zero-to-C2 coverage or proficiency certification. Earlier entries below are historical.

# 3.18.0 — home, choices, plans and wishes

Added Skills 13–16: housing/repair reports, comparisons, future plans/real conditions, and wishes/advice with present subjunctive. New: 32 vocabulary placements, 40 bilingual models, 48 form checks, eight Skills passages / 24 questions, four saved production tasks and 60 recorded future/subjunctive reference slots. Two new Family & Home Word Lab passages add six questions. This closes the previous home-topic transfer gap, not all housing/family language.

Batch media: eight illustrations, 114 SVG cues, 177 recordings / 590.01 seconds. Library: 3,176 indexed clips / 7,442.42 seconds / 45,880,407 bytes. Skills totals: 16 units, 128 vocabulary placements, 144 form checks, 32 passages, 96 questions and 368 review cards. Word Lab: 96 entries unchanged, now four transfer passages / twelve questions.

Route: 1 → 2 → 3 → 4 → 5 → 8 → 9 → 13 → 14 → 6 → 7 → 10 → 11 → 12 → 15 → 16. Stable IDs, original lessons, design, Matrix and eleven JW situations preserved. See LIFE-PLANS-EXPANSION.md for source retrieval, artwork notes, boundaries and maintenance. Broader advanced language, voice diversity and assessed production remain incomplete; zero-to-C2 completion is not claimed. Earlier entries are historical.

# 3.17.0 — four practical units and connected route

Added Skills 9–12: appointments/help, directions/commands, por/para, reservations/repair. New: 32 vocabulary placements, 40 bilingual models, 48 form checks, eight fresh passages / 24 meaning questions, four saved writing/speaking tasks, eight illustrations, 108 distinct SVGs and 169 recorded clips (511.83 seconds). Command reference: 30 regular plus 16 irregular affirmative/negative slots with audio. Skills totals: 12 units, 96 vocabulary placements, 96 form checks, 24 passages, 72 questions. Library: 2,999 indexed clips / 6,852.41 seconds / 42,271,322 bytes.

Suggested route: 1 → 2 → 3 → 4 → 5 → 8 → 9 → 6 → 7 → 10 → 11 → 12. Unit IDs are stable; original lessons show all matching extensions. See PRACTICAL-EXPANSION.md for sources, boundaries and maintenance. Original course, Matrix and eleven JW situations preserved. Remaining broader/advanced coverage and assessed production are not claimed complete. Earlier entries below are historical.

# 3.16.0 — researched foundations additions

Two Skills units added: s7 object-pronoun roles and placement; s8 ser/estar/hay/tener in context. The source review is in CURRICULUM-REVIEW.md and FOUNDATIONS-RESEARCH.md. Added: 16 vocabulary placements, 19 bilingual models, 24 form checks, four fresh passages with twelve meaning questions, two saved production tasks, four original illustrations, 54 distinct SVGs and 66 indexed recordings / 216.02 seconds. Skills totals: eight units, 64 vocabulary placements, 48 form checks, sixteen passages and 48 comprehension questions. Audio total: 2,830 clips / 6,340.58 seconds / 39,134,159 bytes. Original lessons, 96 original illustrations, 11 JW situations and Matrix retained. Broad communicative coverage, teacher-scored production and advanced depth remain incomplete. Earlier notes below are historical.

# 3.15.0 update

JW 4.2 adds a 344-word dialogue, four unique illustrations, twelve distinct vocabulary diagrams, four form checks, six comprehension questions and a fresh listening task. Eleven of 24 situations ready; thirteen planned. Totals: 44 expansion illustrations, 118 vocabulary targets, 228 SVGs. Seventeen new recordings add 227.14 seconds; indexed audio: 2,764 clips / 6,124.56 seconds / 37,811,959 bytes. Original lessons, 96 original illustrations and Matrix retained. Extract the complete ZIP and open verbomaster/index.html; export/import progress on a browser/location change. Earlier entries below are historical.

# 3.14.0 update

JW 4.1 is authored: 189-word beginner dialogue, four unique illustrations, ten vocabulary diagrams, four form checks, six comprehension questions and a fresh listening task. Ten of 24 situations are ready, fourteen planned. Total expansion SVGs: 206. Sixteen new recordings add 154.82 seconds; indexed library: 2,747 clips / 5,897.42 seconds / 36,442,415 bytes. Original lessons, 96 original illustrations and Matrix remain unchanged. Complete ZIP distribution; export/import progress on a location/browser change. Earlier entries below are historical.

# 3.13.0 — chapter 3 complete

JW 3.3 adds a 632-word dialogue, four scene illustrations, fourteen vocabulary targets with distinct conceptual drawings, four form checks, six comprehension questions and a fresh listening message. Twenty-two recordings add 374.01 seconds. Current indexed audio: 2,731 clips / 5,742.60 seconds / 35,507,185 bytes. Nine situations and 36 expansion illustrations are authored; fifteen situations remain planned. Chapter 3 grows 227 → 386 → 632 words. The durable delivery is now the complete project ZIP: extract and open verbomaster/index.html. Export/import progress when moving browser/location; IDs are unchanged. The standalone HTML is a cached diagnostic, not a second durable media copy. Earlier entries below are historical.

# 3.12.0 — preparing and rehearsing a comment

JW situation 3.2 adds a 386-word dialogue, four unique scene illustrations, twelve vocabulary targets with distinct SVG drawings, four form checks, six comprehension questions, a fresh listening message and a writing/revision task. Twenty-one new recordings add 242.63 seconds. Current audio: 2,709 indexed clips / 5,368.59 seconds / 33,254,445 bytes. Eight of 24 situations are authored with 32 scene images; 3.3 and fifteen later situations remain planned. The Next flow, sidebar, design, Matrix and existing progress are preserved. July 2026 source provenance and fiction boundaries are documented in JW-EXPANSION.md. Earlier notes below are historical snapshots.

# 3.11.0 — finding an article and Bible references

New JW situation 3.1 adds a 227-word dialogue, four unique scene illustrations, ten vocabulary targets, four form checks, six comprehension questions and one fresh listening message. Ten new vocabulary SVGs are explanatory drawings; ten other SVGs are answer-neutral form/comprehension cues. Eighteen new recorded clips add 167.76 seconds. Current audio: 2,688 indexed clips / 5,125.96 seconds / 31,791,020 bytes. Seven of 24 JW situations are now authored with 28 scene illustrations; 3.2, 3.3 and fifteen later situations remain planned. Existing Matrix, design, progress and original illustrations are preserved. See JW-EXPANSION.md for the explicitly historical 2019 source and scope. Earlier notes below are historical snapshots.

# 3.10.0 — integrated course flow and extended Matrix

The JW sidebar now nests existing situations/scenes/tasks under their parent, and the bottom Next follows the course sequence. The separate expansion panel is removed. Verb Matrix adds an opt-in five-category view across the existing 20 verbs: 600 added person rows plus 120 accepted -se alternatives, with distinct colors, rules, search, SRS and recordings. Audio adds 604 clips / 691.26 seconds; total 2,670 clips / 4,958.20 seconds / 30,777,420 bytes. Six JW situations and 24 expansion illustrations remain unchanged; 18 situations are planned. The original design and four-tense Studio are retained. See MATRIX-UPDATE.md. Earlier sections are historical snapshots.

# 3.9.1 — chapter 2 complete

Situation 2.3 adds a 556-word Spanish dialogue, four unique scene illustrations, twelve vocabulary targets, four form checks, six comprehension questions, one new listening message, 22 SVG text cues and 21 indexed recordings (343.95 seconds). Chapter 2 progresses 166 → 333 → 556 words. Expansion totals: six authored situations, 24 images, 60 vocabulary targets, 24 form checks, 36 comprehension questions and 120 SVG cues. Eighteen of 24 planned situations remain unauthored. The original design, lessons and progress IDs remain intact. Audio totals: 2,066 indexed clips / 4,266.94 seconds / 26,399,978 bytes. Earlier notes below are historical snapshots.

# 3.9.0 — welcoming people: 2.1 and 2.2

Two complete situations add 166/333-word Spanish dialogues, eight new scene illustrations, 18 vocabulary targets, eight form checks, twelve comprehension questions, two fresh listening messages, 38 SVG text cues and 30 indexed audio clips. Expansion totals: five authored situations, twenty scene images, 48 vocabulary targets, twenty form checks, thirty comprehension questions and 98 cue diagrams. All authored situations have four distinct scene images. Nineteen of 24 situations, including 2.3, remain planned. Audio totals: 2,045 clips / 3,922.99 seconds / 24,328,109 bytes. The one-narrator voice and existing audio remain unchanged. See JW-EXPANSION.md for source provenance and limits. Earlier notes below are historical snapshots.

# 3.8.1 — first expanded JW batch fully illustrated

All twelve scene illustrations for situations 1.1–1.3 are now installed, with four distinct images per situation. The last two show Elena drafting a summary with Ana at home and discussing her preferences with Sara outside. Captions and roadmap status now match the completed artwork. Existing dialogue, vocabulary, recordings and progress IDs are unchanged. The next batch is 2.1–2.3; 21 planned situations remain unauthored. Earlier release notes below describe historical snapshots.

# 3.8.0 — first expanded JW batch

Situations 1.1–1.3 add progressively longer bilingual dialogues (165/307/502 Spanish words), 30 vocabulary entries, 12 form checks, 18 comprehension questions, three new listening messages, 60 text-based SVG cue diagrams, ten scene illustrations and 50 recorded clips. Two illustrations for 1.3 remain pending; 21 further situations are roadmap only. Original 96 illustrations and all prior content remain intact. Audio now totals 2,015 indexed clips / 3,567.48 seconds / 22,183,831 bytes, with automated non-silence checks. This is one recorded Mexican-Spanish narrator, not multiple character voices. See JW-EXPANSION.md for research, scope, tests and maintenance. Earlier entries below are historical release snapshots.

# 3.7.2 — family/home vocabulary update

Added 24 supported entries: eight relationship/household nouns, four adjectives, four location adverbs, four prepositional phrases and four action verbs. Of these, 23 are new explicit strings; abrir recurs with a new model. Added 24 bilingual models, 24 form checks, 48 distinct SVGs and 60 indexed recordings. Word lab totals: 96 entries/models/form checks, 192 word/form SVGs and six existing food-transfer cues.

Overall totals: 466 vocabulary placements / 444 distinct strings, 344 bilingual models, 354 checked targets and 1,965 indexed audio clips. All 96 lesson illustrations and 27 original-lesson manual support passes are retained. No new listening/reading passage was added to the home pack; broader transfer and vocabulary work remain unfinished. Topic filters include all supported words, food/shopping and family/home. Historical notes follow.

# 3.7.1 — food/shopping and transfer update

Added 24 entries (8 nouns, 4 adjectives, 4 adverb/comparison expressions, 4 action verbs, 4 quantity phrases), 24 bilingual models and 24 contextual checks. Four targets recur with new contexts; 20 are new explicit strings. Word lab totals: 72 entries, 72 models, 72 form checks, 144 word/form SVGs plus six answer-neutral comprehension cues. Two new messages add one listening and one reading task, with six meaning questions and first-attempt/support tracking.

Overall: 442 vocabulary placements / 421 distinct strings, 320 bilingual models, 330 checked targets, 14 Skills/Word-lab passages and 42 comprehension questions. Audio: 1,905 indexed clips (+57), still using the existing stored Mexican-Spanish voice. Original 96-image coverage, styling and progress IDs are preserved. Original-lesson manual support remains 27 lessons; 21 original lessons and six Skills units await the deeper pass. These additions do not establish vocabulary sufficiency or CEFR proficiency. See VOCABULARY-COVERAGE.md.

## Earlier release notes

# 3.7.0 update — Word lab vocabulary expansion

Added 48 entries across adjectives, adverbs, prepositions/phrases and everyday nouns, 48 bilingual models, 48 contextual checks and 96 separate semantic/frame SVGs. There are now 418 vocabulary placements / 401 distinct strings overall, 296 bilingual models and 300 checked targets. The indexed audio library is now 1,848 clips (+105). Original lesson manual support remains 27 lessons, with 21 original and six Skills reviews still pending. All 96 lesson illustrations remain installed. See `VOCABULARY-COVERAGE.md` for scope, sources, maintenance and remaining breadth gaps. Historical release details follow.

# Audio update for version 3.4

The indexed library now contains **1,743 recordings** (161 added for the new skills units), **2,611.04 seconds** and **16,340,059 bytes**. The original 11 longer narrations remain separate. All new skills speech is included offline. Listening challenges and listening-review cards use a neutral player label so playback does not reveal the hidden transcript. See SKILLS-UPDATE.md.

---

## Historical version 3.2 implementation notes

# VerboMaster 3.2 — recorded audio everywhere

## What caused the message

Earlier short-word and sentence buttons used the browser's Web Speech API. That API is not guaranteed to have a Spanish voice on a particular device or inside an embedded preview. The narrated story files were independent of it, but the ordinary listen buttons were not.

## What changed

All authored listen buttons now play **stored MP3 recordings**. The app no longer calls browser speech synthesis and does not require users to install voices, speech models, plugins, API keys or a speech service.

**New library:** 1,582 unique short recordings, about 34 minutes 56 seconds of speech, 13,189,885 bytes of MP3 data. This is in addition to the 11 existing full story, specialist lesson and pronunciation overview recordings.

Coverage includes:

- All 224 authored bilingual models' Spanish text, with repeated text reused.
- All 322 vocabulary placements, with identical spoken chunks reused.
- Exercise answer targets and configured Spanish variants.
- Every generated SRS model context and applicable Spanish answer.
- Verb infinitives, conjugated forms, and all 480 person/tense combinations.
- The first lesson's vowel and consonant/digraph examples.
- Spatial-grammar sentences and chunks.
- Original neighborhood story scene buttons.
- Additional Spanish lesson-title recordings, available for future title-listen controls.

English explanations and the English classification answer “ironic criticism” are not presented as Spanish audio. User-written journal text and arbitrary future content do not automatically have a recording; new authored text must pass the audio build/coverage check before release.

## Voice and controls

The short clips use a consistent locally generated neural **Mexican Spanish** voice. It may sound different from the previously selected story narrator. Existing story/lesson narration has been retained, not overwritten.

- Tap any ♪ button to open the compact player.
- Replay, pause, seek or close it.
- Choose Slow / Natural in Settings; this also updates the active recording.
- Starting another recording pauses the previous one.
- Exercises expose answer audio after checking, not before retrieval.
- End-of-lesson mini-cards expose audio with the revealed answer.
- No misleading Spain/Mexico browser-voice switch: the included short-clip locale is explicitly labeled Mexico.

The live project serves local audio files from `assets/speech/`. The portable HTML embeds the same audio as data URLs, so its speech does not depend on network access. The downloadable project ZIP contains the web project and all audio files; the portable HTML is distributed separately to avoid duplicating the entire library in the ZIP.

## Tests completed

`test-audio.mjs` passed with both `speechSynthesis` and `SpeechSynthesisUtterance` deliberately unavailable.

- All 1,582 generated MP3s loaded in Chromium with valid, positive durations.
- Generation checked for nonempty, nonsilent output for each clip.
- Every actual authored `data-speak` button was checked against the audio library across all 48 lessons, the matrix, every verb/tense studio combination, spatial grammar, stories, and all vocabulary/exercise SRS cards.
- Actual playback was verified from a lesson model, exercise answer and consecutive mini-flashcards.
- Slow-speed settings and the mobile player were tested.
- Existing lesson, SRS, persistence and responsive-layout tests continue to pass.
- The final portable HTML was tested with network access disabled: lesson and pronunciation clip playback worked, all 1,582 library entries resolved to embedded audio, and every story image/recording loaded.

These are technical coverage and playback checks, not a claim that a human phonetician listened to and certified every synthesized utterance. Synthesized stress, intonation and uncommon terms can still benefit from a speaker's review. A particular device may require the normal browser permission/user gesture to start media; the native player provides a play control for that situation.

## Build and maintenance

Runtime: static HTML/CSS/JavaScript and MP3 files only. The synthesis software and model are **not** runtime dependencies.

Build tools used:

- Python `piper-tts` 1.8.0, ONNX Runtime and `lameenc`.
- Piper model `es_MX-claude-high` (22,050 Hz; generated as mono MP3 at 48 kbps).
- Length scale 1.08 for a slightly measured pace; small leading/trailing pauses and short inter-sentence gaps.

After updating authored Spanish text:

```sh
# With the local app server running:
node collect-audio.mjs
# Install build dependencies only if regenerating speech:
pip install piper-tts lameenc
# Download the model/config specified in AUDIO-LICENSES.md into
# /home/user/.cache/verbomaster-tts/ (or adapt MODEL in the generator).
python generate_audio_library.py
npm test
python build_portable.py
```

The generator reuses existing clip files, writes stable hash-based filenames, and rebuilds `audio-index.js`. The index maps normalized text to an audio asset. Neither accents nor ñ are removed. The library is deduplicated by spoken text, not by card count.

Records: `audio-inventory.json`, `audio-clips.json`, `audio-build-report.json`, `audio-test-results.json`. Voice provenance: `AUDIO-LICENSES.md`.

## Verified 3.7.0 release
All seven regression suites passed. The portable app and Word lab also passed offline checks. All 48 new contextual answers, 24 lab layouts, 96 new SVGs, indexed audio and actual playback were checked; persisted mistakes and backup round-trip retained their schedules. No reported clipping, overflow, sampled axe violations or runtime errors in the tested states. The 96 lesson illustrations remain intact. These checks do not certify CEFR vocabulary sufficiency or pronunciation quality.

## Verified release 3.7.1
All eight regression suites passed. The base portable app, Word lab and new transfer tasks also passed offline checks. Verified 72 contextual answers, 36 Word-lab layouts, six transfer layouts, all 1,905 indexed audio files, real playback without browser speech, first-attempt preservation, supported-repeat tracking, SRS mistakes and valid/invalid backup handling. No reported overflow, text clipping, sampled axe violations or runtime errors in tested states. All 96 lesson images remain installed. These tests are not pronunciation or CEFR certification.

## Verified release 3.7.2
All eight regression suites passed, followed by the base portable test and both offline Word-lab/transfer suites. Checked all 96 contextual answers, 108 topic/class/phase/viewport combinations, indexed audio and real playback, SRS mistake persistence and backup round trips. No reported overflow, SVG text clipping, sampled axe violations or browser errors in tested states. All 96 lesson illustrations remain installed. These scoped checks are not CEFR or pronunciation certification.