Voice-to-Text Transcription

Wispli's core feature is fast, accurate voice-to-text transcription. Speak naturally and get polished text instantly.

Technology

Wispli uses Whisper Large V3 Turbo powered by Groq for transcription:

• 99%+ accuracy for clear speech
• Near-instant processing (under 1 second)
• 70 languages supported
• Noise-resistant with audio preprocessing

How to Transcribe

Basic Usage

1. Press and hold Ctrl + Space (or your custom shortcut)
2. Speak naturally
3. Release the shortcut
4. Text is transcribed and copied to clipboard

Recording Modes

Mode	How It Works
Hold to Record	Press and hold shortcut, release to stop
Toggle	Press once to start, press again to stop

Change mode in Settings → General → Recording Mode

Audio Processing

Wispli automatically optimizes your audio:

Input Processing

• Echo cancellation - Removes speaker feedback
• Noise suppression - Reduces background noise
• Auto gain control - Normalizes volume levels
• Mono channel - Optimized for speech

Text Cleanup

After transcription, Wispli automatically:

1. Removes filler words ("um", "uh", "like", etc.)
2. Fixes grammar and punctuation
3. Corrects capitalization
4. Removes accidental repetition
5. Applies your chosen formatting style

Transcription Quality

Tips for Best Results

Do	Don't
Speak clearly and naturally	Mumble or speak too fast
Use a good microphone	Use laptop built-in mic in noisy room
Complete your sentences	Stop mid-sentence
Stay within 1 meter of mic	Be too far from microphone

When to Use Custom Vocabulary

If Wispli consistently misrecognizes specific words:

Problem: "React" transcribed as "react" or "re-act"
Solution: Add "React" to Custom Vocabulary

Problem: "Kubernetes" transcribed as "Cooper Netties"
Solution: Add "Kubernetes" to Custom Vocabulary

Learn more about Custom Vocabulary →

Output Options

Automatic Clipboard Copy

By default, transcribed text is copied to clipboard. Just paste with Ctrl + V.

Direct Input Mode

Insert text directly into the active application (coming soon).

History

All transcriptions are saved to History. Access with Ctrl + H or click the clock icon.

Accuracy Confidence

Wispli provides confidence indicators:

Confidence	Meaning
High (green)	Clear speech, confident transcription
Medium (yellow)	Some uncertainty, review recommended
Low (red)	Poor audio quality, may need correction

Handling Edge Cases

Multiple Speakers

Wispli is optimized for single-speaker transcription. For meetings with multiple speakers, consider recording each person separately.

Background Noise

Enable noise suppression in Settings → System → Microphone:

• Light: Minimal processing (quiet environments)
• Medium: Balanced (recommended)
• Heavy: Aggressive filtering (noisy environments)

Accents

Wispli handles most accents well. If accuracy is low:

1. Try speaking slightly slower
2. Enable the language-specific model in Settings

Performance

Metric	Value
Latency	< 1 second
Audio Format	WebM/Opus at 32kbps
Max Recording	5 minutes per session
Offline Mode	Requires API key

Next Steps

• Formatting Styles - 14 ways to format your text
• Translation - Transcribe and translate
• Troubleshooting - Fix common issues