Files
personal-agent-os/memory-systems/agent-os/04_projects/voicestudio/voicestudio-american-accents.md
T
2026-09-16 12:01:21 -07:00

4.2 KiB

title, date, type, status, author, tags, related
title date type status author tags related
American English Accents — VoiceStudio (use) + OpenMAIC (how-to) 2026-09-16 reference active Jimmy
voicestudio
open-maic
accents
tts
american-english
reference
04_projects/veripath/Call-2026-08-15-Tony
04_projects/veripath/tools-inventory

American English Accents

2026-09-16, by Jimmy. Two repos, two mechanisms. VoiceStudio: American accents are dialect-driven (en-US) and are already supported — no code edit required. OpenMAIC: accents are en-US TTS voices picked in the config UI. See the open question at the foot.

VoiceStudio — American English is already supported (no modification needed)

Repo: debpalash/VoiceStudio, adopted 2026-09-16. Cloned locally to C:\Users\fight\Projects\VoiceStudio (commit 8a1e6eb, branch main, 3009 files, no LFS).

Accent in VoiceStudio is a dialect/locale applied on the dub/refine path — not a voice picker:

  • frontend/src/api/dialects.ts:11-20 → DIALECTS.en = ['en-US','en-GB','en-AU','en-IN']; dialectOptionsFor('en') surfaces en-US in the UI dropdown.
  • backend/api/routers/dub_translate.py:94-124 → DIALECT_HINTS["en-US"] = "American English: use US spelling and vocabulary."; dialect_clause() (line 127) injects this hint into the LLM translation prompt, so en-US actually yields US spelling & vocabulary.
  • backend/services/pronunciation.py → Settings → Pronunciation dictionary: respelling rules scoped by language prefix (en-US → prefix en), e.g. Nevada → "Nuh-VAD-uh" (see tests/test_pronunciation.py:170).

How to use American accents: open the dub/translate UI → select dialect en-US → (optional) add custom respellings in Settings → Pronunciation for words you want spoken a specific way. No repository change needed to get American English output.

Note: VoiceStudio does not share OpenMAIC's settings.ts / voice-resolver.ts architecture (searching VoiceStudio for browser-native-tts/ttsVoice returns zero matches). Treating them as the same accent system was the original error — corrected.

OpenMAIC — how to select an American voice

Repo: THU-MAIC/OpenMAIC, cloned to C:\Users/fight\Projects/OpenMAIC (commit 2cbd011, branch main). Accents here = en-US TTS voices.

  1. Open the TTS config popover (toolbar speaker button) → choose a provider → choose its en-US voice. The selection persists in the settings store (ttsVoice + ttsProviderId) and resolveAgentVoice() (lib/audio/voice-resolver.ts) applies it to the teacher narrator and discussion agents.
  2. American voices in the catalog (lib/audio/constants.ts):
    • OpenAI (openai-tts, needs TTS_OPENAI_API_KEY) — alloy, ash, ballad, coral, echo, fable, marin, nova, onyx, sage, shimmer, verse; all tagged language: 'en' (US English).
    • Azure (azure-tts, needs TTS_AZURE_API_KEY + TTS_AZURE_BASE_URL) — en-US-JennyNeural, en-US-GuyNeural and other en-US-…Neural voices (explicit locale).
    • browser-native (no key) — pick a voice whose locale is en-US.
  3. Trap: shipped default is browser-native-tts + 'default' (lib/store/settings.ts:519-520). 'default' means "the browser's first voice" — not guaranteed American. Lock an accent by selecting a real en-US voice instead of 'default'.
  4. Per-agent: resolveAgentVoice() checks overrides[agent_id] first — set a per-agent override in the settings store to force a specific agent (e.g. the Phoenix salon teacher) to always speak American.

VeriPath tooling (context)

This note pairs with [[04_projects/veripath/tools-inventory]], which captures VeriPath's live local services, repo tools, templates, and current status (incl. FreeToken model currently MiniCPM5-2B).

Open question (needs your call)

  • VoiceStudio: make en-US the default dialect (American on without manual selection)? The current default is no dialect → neutral English (dialect_clause returns "" when unset). If yes, I'll locate the default-dialect site and propose the exact line for your approval before any edit.

Author: Jimmy. Status: reference; revisit when the default-dialect decision is made.