Piper

Piper Text-to-Speech Workflow

Convert text or markdown files to natural-sounding speech audio.

Parameters

Available Voices

Models are stored in ~/piper-voices/. Each voice requires two files: .onnx and .onnx.json.

Download additional voices from: https://huggingface.co/rhasspy/piper-voices

Workflow

Step 1: Validate Input

Determine input type and validate:

# Check if input is a file
if [[ -f "$input" ]]; then
  input_type="file"
  extension="${input##*.}"
else
  input_type="text"
fi

Constraints:

• If file: must exist and be readable
• If markdown (.md): proceed to Step 2
• If text file (.txt) or literal text: skip to Step 3

Step 2: Clean Markdown (if applicable)

For markdown files, clean formatting for TTS:

python3 /Users/katiemulliken/Documents/Projects/kmtools/piper/clean_obsidian_for_tts.py "$input" -o "${input%.md}_clean.txt" --stats

Removes:

• YAML frontmatter
• Markdown formatting (headers, bold, italic, links)
• Code blocks and inline code
• HTML tags and comments
• Emojis and special Unicode
• URLs

Output: Cleaned text file for Step 3

Step 3: Generate Audio

Convert text to speech using Piper:

# From cleaned/text file
/Users/katiemulliken/.local/bin/piper \
  -m ~/piper-voices/${voice}.onnx \
  -i "$cleaned_input" \
  -f "$output" \
  --length-scale $speed \
  --sentence-silence $sentence_silence \
  --volume $volume

# From literal text
echo "$text" | /Users/katiemulliken/.local/bin/piper \
  -m ~/piper-voices/${voice}.onnx \
  -f "$output" \
  --length-scale $speed \
  --sentence-silence $sentence_silence \
  --volume $volume

Step 4: Verify Output

Confirm successful generation:

# Check file exists and has content
if [[ -f "$output" && -s "$output" ]]; then
  echo "Audio generated: $output"
  # Get duration using ffprobe if available
  ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$output" 2>/dev/null
fi

Step 5: Link Audio to Original

You should link the audio to the original input using the following syntax

![[The Converted Audio File.wav]]

Complete Examples

Markdown to Audio (Default Settings)

# Step 1-2: Clean markdown
python3 /Users/katiemulliken/Documents/Projects/kmtools/piper/clean_obsidian_for_tts.py "article.md" -o "article_clean.txt" --stats

# Step 3: Generate audio
/Users/katiemulliken/.local/bin/piper \
  -m ~/piper-voices/en_US-amy-medium.onnx \
  -i "article_clean.txt" \
  -f "article.wav" \
  --length-scale 0.67 \
  --sentence-silence 0.3

Text to Audio (One-liner)

echo "Hello, this is a test." | /Users/katiemulliken/.local/bin/piper \
  -m ~/piper-voices/en_US-amy-medium.onnx \
  -f "output.wav"

Batch Process Directory

for file in *.md; do
  python3 /Users/katiemulliken/Documents/Projects/kmtools/piper/clean_obsidian_for_tts.py "$file" | \
    /Users/katiemulliken/.local/bin/piper \
      -m ~/piper-voices/en_US-amy-medium.onnx \
      -f "${file%.md}.wav" \
      --length-scale 0.67 \
      --sentence-silence 0.3
done

Error Handling

Adding Voices

cd ~/piper-voices
# Download model and config (example: lessac voice)
curl -L "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/en_US-lessac-medium.onnx" -o en_US-lessac-medium.onnx
curl -L "https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/en_US-lessac-medium.onnx.json" -o en_US-lessac-medium.onnx.json

Preview voices: https://rhasspy.github.io/piper-samples/