Piper TTS Swedish Voices for NVDA (Sonata)

High-quality Swedish neural TTS voices for NVDA screen reader via the Sonata addon.

• Gender: Female
• Quality: Medium (22050 Hz)
• Size: 56 MB
• Training data: NST Swedish speech corpus
• License: CC-BY 4.0

1. Install NVDA (2024.1 or later)
2. Install the Sonata Neural Voices addon

1. Download the voice package (.tar.gz) from Releases
2. In NVDA, open Tools → Sonata Voice Manager
3. Click Install from local file
4. Select the downloaded .tar.gz file
5. Restart NVDA
6. Go to Settings → Speech and select Sonata Neural Voices
7. Choose Alma from the voice list

You can test the voice with Piper directly:

echo "Hej! Det här är den svenska rösten Alma." | piper --model sv_SE-alma-medium.onnx --output_file test.wav

These voices are trained using Piper TTS and optimized for screen reader use. They provide significantly better quality than the default eSpeak NG synthesizer for Swedish.

Created by Daniel Nylander to improve accessibility for Swedish-speaking visually impaired users.

Voice models: CC-BY 4.0 This repository: GPL v2 (matching NVDA and Sonata licenses)

• Gender: Male
• Quality: Medium (22050 Hz)
• Size: 80 MB
• Training data: Voice cloned from 3 minutes of speech
• License: CC-BY 4.0

• Gender: Male
• Quality: Medium (22050 Hz)
• Size: 63 MB
• Training data: NST Swedish speech corpus
• License: CC-BY 4.0