Documentation

System requirements

• Windows 10 or Windows 11, 64-bit.
• At least 2.5 GB free for installation and temporary setup files.
• Additional model storage from approximately 75 MB to 3.1 GB.
• A CPU is sufficient; a compatible NVIDIA CUDA device accelerates supported profiles.
• Internet access for the first download of each transcription model.

Installation

1. Download Qevoxyl-2.0.2-Setup.exe.
2. Close any running Qevoxyl window.
3. Run the installer and complete the per-user setup.
4. Open Qevoxyl from the Windows Start menu.

Start a new transcription

1. Select New transcription from Home or the completed transcript view.
2. Choose an audio/video file or open the recorder.
3. Review language, model, device, and advanced options.
4. Start the job. Qevoxyl displays only the active stage: Ready, Running, or Transcript.

Supported inputs include MP3, WAV, M4A, AAC, FLAC, OGG, OPUS, WMA, MP4, MKV, MOV, AVI, and WEBM.

Record audio

The Qevoxyl recording modal displays the selected device, recording duration, and live input level. Use the microphone selector before starting when a USB microphone, audio interface, or other external device is connected. Stop saves and selects the recording; Cancel discards the current capture.

Processing and queue

The Running screen reports transcription percentage, current segment, elapsed audio, CPU and RAM use, engine plan, and processing speed. Completed segments are checkpointed so supported interrupted work can resume.

Multiple files are processed sequentially. The queue groups the active job with files waiting behind it. You can pause or cancel active work and remove jobs that have not started.

Transcript workspace

• Play the source, jump backward/forward five seconds, and change playback speed.
• Switch between timestamped segments and continuous full text.
• Edit text with autosave, undo/redo, find, replace, and replace all.
• Review uncertain segments and update detected speaker names.
• Create a cleaned copy or queue local audio-to-English translation.

Transcript versions

• Original — the editable source transcript; always preserved.
• Cleaned — a non-destructive copy with optional filler and repetition removal.
• English — English-only local Whisper translation.
• Bilingual — original and English text paired together.

Models and performance

The Models page manages Tiny, Base, Small, Medium, and Large-v3. It shows download status, disk use, quality guidance, and whether a model fits the detected hardware. Settings offers Eco, Balanced, Performance, and Maximum Accuracy profiles as well as idle model-unload timing.

Export formats

Export the Original, Cleaned, English, or Bilingual version in timestamped or full-text form.

• TXT for general plain text.
• SRT / VTT for subtitles and web captions.
• JSON for structured transcript data.
• DOCX / PDF for finished documents with the Qevoxyl logo and watermark.

Storage, history, and logs

Qevoxyl stores application data under %LOCALAPPDATA%\Qevoxyl\, including its database, models, recordings, exports, and logs. History and the in-app log viewer support pagination with 20, 50, or 100 rows per page.

Settings > About displays the application version and exact build number. Include both values when reporting a problem.