Pronunciation Comparison Studio
Use your own authorized reference clip and record yourself, or upload an existing take. Choose the start of each clip and play equal-length A/B windows repeatedly. Inspect waveform energy and timing, then write what you hear. This tool does not transcribe, recognize phonemes or grade an accent.
Key features
- Reference file and microphone recording or own-file input kept in the browser
- Explicit microphone request only after pressing Record, with stop/cancel and track cleanup
- Independent reference, take and A-then-B listening within equal-length selected windows
- Side-by-side waveform envelopes and approximate high-energy start/length differences
- Self-assessment and note exported as JSON without embedded audio or a machine pronunciation score
- A clearly labeled synthetic tone demo that needs no microphone
How to use
- Choose a short reference audio file you are allowed to use.
- Press Record to request microphone permission and capture your take, or upload a take file instead.
- Adjust the reference and take start times and common window length, then play either clip or A then B repeatedly.
- Use the waveforms and rough timing numbers as listening aids, not as phonetic or native-speaker judgments.
- Write a self-review and download the JSON report; download your recording separately if desired.
Use cases
- Compare your own reading of a short phrase with an authorized model clip
- Repeat equivalent A/B windows while practicing timing and rhythm
- Keep a personal practice note without sending audio to a service
Frequently asked questions
Does the app score how native my accent sounds?
No. It does not recognize words or phonemes, measure intelligibility or produce a native-speaker score. The high-energy span is only an approximate amplitude cue and can include breaths, noise or music. Judge pronunciation by listening and, when needed, a qualified person.
When is microphone permission requested?
Only when you press Record on a secure page. You may instead upload an audio take. Cancel, stop, clear, leaving the page and component cleanup stop microphone tracks. The browser or operating system can deny access.
What does the alignment do?
You set a separate start time for each clip and one common playback duration. A and B play sequentially over those windows without changing pitch or creating a merged file. The highlighted waveform window follows these values.
How are the timing differences calculated?
The duration difference is the take's decoded duration minus the reference's. The high-energy start and length differences come from 96 peak-amplitude bins using a threshold relative to each clip's peak. They are approximate and not a speech detector or quality grade.
Where is the audio stored?
Decoded clips and a recorded blob stay in this tab. The JSON report has numbers and your note but no audio. A recording download is separate. No clip is uploaded by this tool; closing or clearing releases local resources.
Privacy
The tool decodes and plays your audio locally. Microphone access is requested only after your Record click. It does not upload clips. You decide whether to download a report or recorded take; confirm your right to use a reference file.
Comments & questions