None of 13 additional open-source speaking coaches documents all five strict capabilities

Asked (summary)

Find, list and factually compare all fully open-source, ready-to-deploy all-in-one apps for spoken-English practice with AI coaching, ASR/error detection, corrective practice, progress tracking, drills and conversation; document facts and explain exclusions.

Asked (follow-up)

Find another open-source project for English language practice featuring pronunciation analysis, error detection, active correction, forced speaking, and progress tracking. Crucially, the AI must be capable of genuinely recognizing errors in pronunciation, sentence structure, word choice, and so on.

New projects / stricter verification: this round adds 13 candidates checked against official repository and package-index documentation on 2026-09-12. A capability counts only when the README documents a genuine mechanism — plain speech-to-text, transcript matching, an ASR-confidence score, a mode named "Pronunciation Coach", or an LLM judging a transcript does not count as pronunciation analysis. The strict all-five column is false for every candidate; no finding below softens that.

Capability matrix: documented vs partial vs absent

documented genuine mechanism partial: claimed, weak, or not exposed to the user not documented / not present

What the matrix cannot say alone

mcp-server-pronunciation is the only candidate with a genuine acoustic engine plus spoken grammar feedback plus an explicit retry loop — it fails only on persistent progress, and its README calls the pronunciation signals experimental. github.com

Oral English Practice has the strongest documented forced-remediation and progress loop — recurring mistakes re-drilled until clean, every session in data.csv — but no acoustic pronunciation engine at all, and it depends on the Claude app rather than standing alone. github.com

LinguaCompanion has Azure phoneme scoring wired in the backend, but its README states it is not yet exposed in the UI, so it does not presently satisfy the pronunciation requirement as an end-user feature. github.com

LiltLab pairs a genuine wav2vec2 phoneme and prosody engine with saved difficult-word history, but its CC BY-SA 4.0 license is not OSI-approved software licensing, placing it outside the strict open-source-software list. github.com

Project cards

All 13 candidates, row by row

Project	License	Form	Engine as documented	Missing strict requirements	All five	Source

Sources

Method: 13 additional candidate projects (one row each), verified against official GitHub READMEs and PyPI / marketplace pages, checked 2026-09-12. Each cell records whether the repository documents a genuine mechanism for that capability: acoustic/phoneme pronunciation analysis means phoneme scoring or alignment (wav2vec2 CTC, Azure pronunciation assessment, SpeechAce, Praat prosody) — never ASR text matching, confidence scores, or an LLM judging a transcript; grammar must be diagnosed on the learner's spoken response; forced speaking requires an explicit retry/re-drill/mastery workflow. "Partial" marks claimed, weak, or non-exposed implementations. Evidence quotes were trimmed for space; full details sit in each linked repository. The strict all-five-capabilities flag is false for every row.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.