Add voice genders, fix MaryTTS /voices API

Makefile

@@ -1,6 +1,6 @@
 SHELL := bash
 
-.PHONY: check clean reformat dist docker amd64 index test
+.PHONY: check clean reformat dist docker index test genders
 
 all: dist
 
@@ -24,3 +24,6 @@ docker:
 
 index:
 	bin/make_sample_html.py local/ > index.html
+
+genders:
+	scripts/get-genders.sh > larynx/VOICE_GENDERS

larynx/VERSION

@@ -1 +1 @@
-1.0.2
+1.0.3

larynx/VOICE_GENDERS

@@ -0,0 +1,51 @@
+ru-ru_nikolaev-glow_tts M
+ru-ru_hajdurova-glow_tts F
+ru-ru_minaev-glow_tts M
+en-us_scottish_english_male-glow_tts M
+en-us_northern_english_male-glow_tts M
+en-us_cmu_ljm-glow_tts F
+en-us_cmu_slp-glow_tts F
+en-us_cmu_bdl-glow_tts M
+en-us_cmu_aew-glow_tts M
+en-us_cmu_clb-glow_tts F
+en-us_cmu_jmk-glow_tts M
+en-us_cmu_aup-glow_tts M
+en-us_cmu_lnh-glow_tts F
+en-us_southern_english_male-glow_tts M
+en-us_cmu_rxr-glow_tts M
+en-us_blizzard_fls-glow_tts F
+en-us_blizzard_lessac-glow_tts F
+en-us_glados-glow_tts F
+en-us_cmu_fem-glow_tts M
+en-us_harvard-glow_tts F
+en-us_judy_bieber-glow_tts F
+en-us_cmu_rms-glow_tts M
+en-us_cmu_slt-glow_tts F
+en-us_mary_ann-glow_tts F
+en-us_ljspeech-glow_tts F
+en-us_cmu_ksp-glow_tts M
+en-us_cmu_ahw-glow_tts M
+en-us_kathleen-glow_tts F
+en-us_cmu_eey-glow_tts F
+en-us_southern_english_female-glow_tts F
+en-us_ek-glow_tts F
+de-de_hokuspokus-glow_tts F
+de-de_karlsson-glow_tts M
+de-de_eva_k-glow_tts F
+de-de_rebecca_braunert_plunkett-glow_tts F
+de-de_thorsten-glow_tts M
+de-de_pavoque-glow_tts M
+de-de_kerstin-glow_tts F
+nl_nathalie-glow_tts F
+nl_rdh-glow_tts M
+nl_flemishguy-glow_tts M
+nl_bart_de_leeuw-glow_tts M
+sw_biblia_takatifu-glow_tts M
+it-it_riccardo_fasol-glow_tts M
+it-it_lisa-glow_tts F
+fr-fr_siwis-glow_tts F
+fr-fr_gilles_le_blanc-glow_tts M
+fr-fr_tom-glow_tts M
+es-es_carlfm-glow_tts M
+es-es_karen_savage-glow_tts F
+sv-se_talesyntese-glow_tts M

larynx/server.py

@@ -35,9 +35,11 @@
     VOCODER_DIR_NAMES,
     VOCODER_QUALITY,
     VOICE_DOWNLOAD_NAMES,
+    VOICE_GENDERS,
     download_voice,
     get_voices_dirs,
     load_voices_aliases,
+    load_voices_genders,
     valid_voice_dir,
 )
 from larynx.wavfile import write as wav_write
@@ -504,11 +506,49 @@ async def api_process():
 @app.route("/voices", methods=["GET"])
 async def api_voices():
     """MaryTTS-compatible /voices endpoint"""
+
+    load_voices_genders()
+
+    # [voice] [language] [gender] [tech=hmm]
     lines = []
-    for voice_id in get_voices():
-        lines.append(voice_id)
 
-    return "\n".join(lines)
+    # Search for downloaded voices/vocoders
+    for voices_dir in voices_dirs:
+        if not voices_dir.is_dir():
+            continue
+
+        # <LANGUAGE>/<VOICE>-<TTS_SYSTEM>
+        for lang_dir in voices_dir.iterdir():
+            if (not lang_dir.is_dir()) or (lang_dir.name in VOCODER_DIR_NAMES):
+                continue
+
+            # Voice
+            voice_lang = lang_dir.name
+            for voice_model_dir in lang_dir.iterdir():
+                if not valid_voice_dir(voice_model_dir):
+                    continue
+
+                voice_name_tts = voice_model_dir.name
+                full_voice_name = f"{voice_lang}_{voice_name_tts}"
+                voice_name, tts_system = voice_name_tts.split("-", maxsplit=1)
+
+                voice_gender = "NA"  # not available
+                gender_path = voice_model_dir / "GENDER"
+
+                if gender_path.is_file():
+                    # Use GENDER file in voice directory
+                    voice_gender = gender_path.read_text().strip()
+                else:
+                    # Use value from VOICE_GENDERS file
+                    voice_gender = VOICE_GENDERS.get(full_voice_name, voice_gender)
+
+                # Repeat voice for each quality level
+                for quality in VOCODER_QUALITY:
+                    lines.append(
+                        f"{voice_name};{quality} {voice_lang} {voice_gender} {tts_system}"
+                    )
+
+    return "\n".join(sorted(lines))
 
 
 @app.route("/version", methods=["GET"])

larynx/utils.py

@@ -36,6 +36,9 @@
 # voice -> <name>.tar.gz
 VOICE_DOWNLOAD_NAMES: typing.Dict[str, str] = {}
 
+# voice -> gender
+VOICE_GENDERS: typing.Dict[str, str] = {}
+
 
 def load_voices_aliases():
     """Load voice aliases from VOICES file"""
@@ -56,6 +59,20 @@ def load_voices_aliases():
                 VOICE_DOWNLOAD_NAMES[full_voice_name] = download_name
 
 
+def load_voices_genders():
+    """Load genders from VOICE_GENDERS file"""
+    if not VOICE_GENDERS:
+        # Load voice genders
+        with open(_DIR / "VOICE_GENDERS", "r", encoding="utf-8") as voices_file:
+            for line in voices_file:
+                line = line.strip()
+                if not line:
+                    continue
+
+                full_voice_name, gender = line.split(" ", maxsplit=1)
+                VOICE_GENDERS[full_voice_name] = gender
+
+
 def resolve_voice_name(voice_name: str) -> str:
     """Resolve voice name using aliases"""
     load_voices_aliases()

local/de-de/eva_k-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/de-de/hokuspokus-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/de-de/karlsson-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/de-de/kerstin-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/de-de/pavoque-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/de-de/rebecca_braunert_plunkett-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/de-de/thorsten-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/blizzard_fls-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/blizzard_lessac-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_aew-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_ahw-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_aup-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_bdl-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_clb-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_eey-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_fem-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_jmk-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_ksp-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_ljm-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_lnh-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_rms-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_rxr-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/cmu_slp-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/cmu_slt-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/ek-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/glados-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/harvard-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/judy_bieber-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/kathleen-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/ljspeech-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/mary_ann-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/northern_english_male-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/scottish_english_male-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/en-us/southern_english_female-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/en-us/southern_english_male-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/es-es/carlfm-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/es-es/karen_savage-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/fr-fr/gilles_le_blanc-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/fr-fr/siwis-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/fr-fr/tom-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/it-it/lisa-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/it-it/riccardo_fasol-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/nl/bart_de_leeuw-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/nl/flemishguy-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/nl/nathalie-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/nl/rdh-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/ru-ru/hajdurova-glow_tts/GENDER

@@ -0,0 +1 @@
+F

local/ru-ru/minaev-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/ru-ru/nikolaev-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/sv-se/talesyntese-glow_tts/GENDER

@@ -0,0 +1 @@
+M

local/sw/biblia_takatifu-glow_tts/GENDER

@@ -0,0 +1 @@
+M

scripts/get-genders.sh

@@ -0,0 +1,23 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Directory of *this* script
+this_dir="$( cd "$( dirname "$0" )" && pwd )"
+src_dir="$(realpath "${this_dir}/..")"
+
+local_dir="${src_dir}/local"
+
+find "${local_dir}" -mindepth 2 -maxdepth 2 -type d | \
+    while read -r voice_dir;
+    do
+        voice_lang="$(dirname "${voice_dir}" | xargs basename)"
+        if [ "${voice_lang}" = 'hifi_gan' ] || [ "${voice_lang}" = 'waveglow' ]; then
+            continue;
+        fi
+
+        voice_name="$(basename "${voice_dir}")"
+        voice_gender="$(cat "${voice_dir}/GENDER")"
+
+        full_voice_name="${voice_lang}_${voice_name}"
+        echo "${full_voice_name}" "${voice_gender}"
+    done