Persian TTS listening checklist — original VowelMarks diagnostic, October 1, 2026

This is a worksheet, not a scored benchmark. Read the guide's intended readings first.
Use the same source text for every voice. Do not put these instructions in the TTS input.

Tool / voice:
Tier / settings / direction:
Generation date:
Reviewer and Persian variety:
Audio filename:

For each line, record: reading correct / wrong / uncertain; delivery notes; replay time.

01. Ezafe chain: ketab-e jadid-e Narges; no added link between ast and another sentence.
02. Vowel-ending Ezafe: khane-ye kuchak-e ma.
03. گل: gol, flower; gol-e ghermez.
04. گل: gel, mud; not gol.
05. مرد: mard, man.
06. مرد: mord, died.
07. شیر: shir, milk.
08. شیر: shir, lion. The sound is unchanged; do not require a pronunciation difference.
09. Names: Narges, Bahram, Tehran, Shiraz; keep the intended names.
10. Clock: 8:30; hasht-o nim and hasht-o si daghighe are acceptable time readings.
11. Price: sad-o bist-o panj hezar toman; not 125 toman or 12.5 hezar toman.
12. Room: 204; devist-o chahar or do-sefr-chahar, according to the chosen room-number convention.
13. Mixed script: PDF and Wi-Fi remain recognizable; allow intelligible Persian-accented or English forms.
14. Formal verbs: preserve the negative nemi-ravam and future-context mi-ayam; do not drop ne-.
15. Casual register: ketabet and ro; do not require formal grammar in a deliberately casual sentence.
16. Quotation: no spoken quotation-mark names, duplicated dialogue or missing final action.

Second listen — delivery (separate from pronunciation):
Pace / intelligibility:
Pauses and phrasing:
Clipped starts / missing words / repeated words:
Voice consistency:
Suitable for the actual lesson, video or app:

Do not count an unfamiliar accent or uncertain judgment as a confirmed error.
Add three sentences from your own project and document their intended readings.