Free online AMR to RTF converter

A phone voice memo or voicemail stored as AMR must be transcribed before it can be edited or shared as an RTF document. The conversion therefore has two stages: speech recognition creates the words, then the transcript is saved with RTF formatting.

What the AMR format is

AMR (Adaptive Multi-Rate) is a lossy speech-audio format designed for low-bandwidth voice. AMR-NB commonly stores narrowband telephone speech, while AMR-WB stores wider-band speech. Mobile phones, older Android voice recorders, 3GPP-based systems, voicemail platforms, and some telephony applications create .amr files.

An AMR file contains recorded sound, not selectable words. Renaming its extension or changing its container cannot produce a meaningful RTF document; speech-recognition software must analyze the audio first.

What the RTF format is

RTF (Rich Text Format) is a document format for plain text with portable formatting such as fonts, paragraphs, bold and italic text, and basic spacing. Microsoft Word, WordPad, LibreOffice Writer, and many document-management systems can open it.

RTF is suitable when a transcript needs editing, annotation, printing, or exchange between word processors without requiring the original word-processing application. It stores the recognized transcript, not the AMR recording or a guaranteed word-for-word representation of the speech.

How to actually convert AMR to RTF

Using the built-in online converter

1. Open the AMR-to-RTF converter on 101convert.com; the service is available free without installing software.
2. Upload the .amr recording and choose RTF as the output format.
3. If transcription options are provided, select the spoken language and enable speech recognition. Review any available speaker, punctuation, or timestamp settings.
4. Start the conversion and download the resulting .rtf file.
5. Open the result in LibreOffice Writer or Microsoft Word and correct names, numbers, punctuation, and technical terms against the recording.

Using desktop tools

A local workflow gives more control over the audio and recognition model. FFmpeg can decode AMR, Whisper can transcribe the decoded speech, and LibreOffice can package the transcript as RTF:

1. Install FFmpeg, an installed Whisper command-line package, and LibreOffice.
2. Decode the recording to a recognition-friendly WAV file:

ffmpeg -i input.amr -ar 16000 -ac 1 speech.wav

3. Transcribe the WAV file, replacing the language code as needed:

whisper speech.wav --model small --language en --output_format txt --output_dir transcript

4. Convert the generated plain-text transcript to RTF:

libreoffice --headless --convert-to rtf --outdir transcript transcript/speech.txt

For difficult audio, use a larger Whisper model if the computer has sufficient memory, and specify the correct language instead of relying on automatic detection. The final file will normally be named speech.rtf in the transcript directory.

Quality and compatibility limitations

AMR is optimized for speech at low bit rates, so background noise, overlapping speakers, accents, clipped recordings, and telephone-bandwidth audio can reduce transcription accuracy. Lossy compression cannot be reversed before recognition.

RTF has no standard way to preserve reliable word-level timestamps, speaker identification, or confidence scores. Keep the original .amr file alongside the RTF transcript when the recording may be needed for verification. RTF also does not embed the audio unless a particular word processor adds a proprietary attachment.