Run the following command (This should be executed while in the cloned language-resources folder)
export FESTIVAL_SUIT_PATH=<your desired path to install Festival>
mkdir ${FESTIVAL_SUIT_PATH} && ./third_party/festvox/setup.sh
- Make sure to add the environment variables to .bashrc
- Lexicon -
bazel build //bn/festvox:make_lexicon_scm - Festival prompts -
bazel build //bn/festvox:make_festvox_prompts
- Download recordings from
export WAV_FOLDER=<your desired path to store downloaded wav files>
mkdir -p ${WAV_FOLDER} && wget http://storage.googleapis.com/voice-builder-public-data/all_example_data/data/bn_bd_male/bn_bd_male_16.tar.gz && tar xzf bn_bd_male_16.tar.gz -C ${WAV_FOLDER}
- Train TTS Voice (This should be executed while in the cloned language-resources folder). This step will take couple of hours to finish.
export OUTPUT_VOICE_FOLDER=<your desired path to store output voice>
./festival_utils/build_festvox_voice.sh ${WAV_FOLDER} bn ${OUTPUT_VOICE_FOLDER}
- Synthesize one example sentence (This always needs to run from the produced output voice directory).
cd ${OUTPUT_VOICE_FOLDER}
echo 'অলিম্পিক ইন্ডাস্ট্রিজ জীবনকে সহজ করে তুলেছে' |
${FESTIVALDIR}/bin/text2wave \
-eval festvox/goog_bn_unison_cg.scm \
-eval '(voice_goog_bn_unison_cg)' \
> example.wav