Motivational Speech Synthesis

Text-to-motivational-speech with adjustable motivational factor to control motivational prosody.

Preliminary Paper | Project Page | Colab Demo

---

Motivational speech has emerged as a popular audiovisual phenomenon within Western subcultures, conveying optimal strategies and principles for success through expressive, high-energy delivery. The present paper artistically explores methods for synthesizing the distinctive prosodic patterns inherent to motivational speech, while critically examining its sociocultural foundations. Drawing on recent advances in emotion-controllable text-to-speech (TTS) systems and speech emotion recognition (SER), we employ deep learning models and frameworks to replicate and analyze this genre of speech. Within our proposed architecture, we introduce a one-dimensional motivational factor derived from high-dimensional emotional speech representations, enabling the control of motivational prosody according to intensity. Situated within broader discourses on self-optimization and meritocracy, Motivational Speech Synthesis contributes to the field of emotional speech synthesis, while also prompting reflection on the societal values embedded in such mediated narratives.

Cloning

Use --recurse-submodulesflag to also clone submodules

git clone --recurse-submodules git@github.com:MotivationalSpeechSynthesis/motivational-speech-synthesis.git

If using HTTPS rather than SSH for cloning

git clone git@github.com:MotivationalSpeechSynthesis/motivational-speech-synthesis.git
cd motivational-speech-synthesis
git config submodule.emoknob.url https://github.com/tonychenxyz/emoknob.git
git submodule update --init --recursive

Requirements

• Linux OS recommended (Windows support expected but not tested, macOS currently unsupported)

Installation and Running

Note: Each standalone script execution recompiles the model. For repeated experiments and faster iteration, use the provided Jupyter notebook.

Using uv

Run script

uv run motivationalTTS.py "Every journey begins with a single step."

Virtual env for jupyter-notebook:

uv venv

Start jupyter-notebook

uv run jupyter-notebook

Using pip

python -m venv env
source env/bin/activate
pip install -r requirements.txt

Run script

python motivationalTTS.py "Every journey begins with a single step."

Start jupyter-notebook

uv run jupyter-notebook

Optional Parameters

You can customize the synthesis with the following optional arguments:

uv run motivationalTTS.py "Every journey begins with a single step." \
    --motivational-factor 0.8 \
    --seed 42 \
    --intermediate-dir "./output_audio" \
    --output-name "my_audio.wav" \
    --device "cuda:0" \
    --dtype "float16" \
    --debug \
    --average-speaker-emb-dir "average-speaker-embeddings/average-speaker-embeddings_400"

Google Colab

The model can also be run with following Google Colab example