{% extends "layout_noprofile.html" %} {% block body %}

Download Profile

voice2json did not find a profile at {{ profile_path }}

You can either manually download a profile and extract it to that location (so {{ profile_path / "profile.yml" }} exists), or follow the instructions below.


Choose a Profile

Profiles are listed by language and model type with performance characteristics. More ★ means more accurate, higher (Nx) means faster. See testing conditions and the legend for more details.

{% for language in sorted(profiles) %} {% for locale in sorted(profiles[language]) %} {% for name, info in sorted(profiles[language][locale].items()) %} {% endfor %} {% endfor %} {% endfor %}
{{ language }} Locale Description Closed Open
{% set download_path = info.download_path(name, download_dir) %} {% if download_path.exists() %} {% else %} {% endif %} {{ locale }} {{ info.description }} {% if info.accuracy_closed > 0 %} {% for i in range(info.accuracy_closed) %} ★ {% endfor %} {% else %} ☹ {% endif %}   ({{ info.speed_closed }}x) {% if info.accuracy_open > 0 %} {% for i in range(info.accuracy_open) %} ★ {% endfor %} {% else %} ☹ {% endif %}   ({{ info.speed_open }}x)

Testing Conditions

Each profile is given a ★ rating, indicating how accurate it was at transcribing a set of test WAV files. Anything below 75% accuracy is considered unusable (☹).

The closed condition tested the profile on sentences known ahead of time, whereas open used a general language model. The average transcription time relative to the real-time duration of the test WAV files is given as (Nx).

Legend

Transcription Accuracy  
★ ★ ★ ★ ★ [95%, 100%]
★ ★ ★ ★ [90%, 95%)
★ ★ ★ [85%, 90%)
★ ★ [80%, 85%)
[75%, 80%)
[0%, 75%)
{% endblock %}