Skip to content

ConvertPraatToLpc.md

Latest commit

 

History

History
104 lines (76 loc) · 4.87 KB

ConvertPraatToLpc.md

File metadata and controls

104 lines (76 loc) · 4.87 KB

ConvertPraatToLpc

Summary

Converts a pair of speech files in Praat Pitch format and in Praat LPC format to a file in our binary LPC format. Our LPC format is optimized for replay with the TMS52xx speech synthesis chips.

Praat (which is Dutch for "talk") is a powerful phonetics program by the Phonetic Sciences group of the University of Amsterdam. It offers many techniques and algorithms to analyze and synthesize speech. You can operate it from a GUI or from scripts. It's available for the major platforms. For example, on Debian/Ubuntu:

sudo apt install praat

This introduction is a good place to start. You can perform a pitch analysis and an LPC analysis with Praat and then convert the results to our LPC format with this tool.

Usage

java ConvertPraatToLpc [options...] input.Pitch input.LPC unvoiced_energy_factor voiced_energy_factor output.lpc

where:
input.Pitch The pitch input file in the short text format produced by Praat.
input.LPC The LPC input file in the short text format produced by Praat.
unvoiced_energy_factor A scaling factor for the energy of unvoiced frames. A value around 0.5 may work reasonably well.
voiced_energy_factor A scaling factor for the energy of voiced frames. A value around 0.5 may work reasonably well.
output.lpc The output file in our raw binary LPC format.

The options are:
-chip name Target the specified speech synthesis chip: one of TMS5100, TMS5110A. TMS5200 (the default), or TMS5220.
-tms5200 Short for -chip TMS5200 (the default).
-tms5220 Short for -chip TMS5220.
-8kHz Set a sample target rate of 8 kHz (the default).
-10kHz Set a sample target rate of 10 kHz.
-addstopframe Add a stop frame to the LPC frames in the output.

Example workflow

The video tools contain a Praat script that reads a specified WAV file, analyzes the speech, and writes a Praat pitch file and a Praat LPC file. You can follow it interactively or execute it from the command-line:

praat --run tools/lpc.praat \
  /tmp/speech.wav \
  /tmp/speech.PraatPitch \
  /tmp/speech.PraatLPC \
  250 550 0.02 0.40 0.20 0.20 0.03

The header of the script explains the parameters. You can then apply the conversion tool:

java ConvertPraatToLpc \
  -addstopframe \
  /tmp/speech.PraatPitch \
  /tmp/speech.PraatLPC \
  0.4 0.6 \
  speech.lpc

You can convert the LPC coefficients to a WAV sound file, in order to quickly check the quality of the speech in any audio player:

java ConvertLpcToWav -tms5200 -analog -precise speech.lpc check.wav

We're choosing the common analog dynamic range, but with a full 16-bit precision, so we can identify any artifacts more easily.

If you're happy with the results, you can embed the LPC speech in an executable cartridge for the TI-99/4A. First create a TMS video file, in this example only containing the speech:

java ComposeVideo -ntsc speech.lpc speech.tms

Then package the video file with a small default player program in a cartridge:

java PackageVideoInCartridge -title 'MY SONG' speech.tms speech.rpk

Instead of an RPK file speech.rpk for the Mame emulator, you can also create just a raw ROM file like romc.bin, for other emulators.

You can then run the application from a RAM cartridge on the computer or in an emulator like Mame:

mame ti99_4a -ioport peb -ioport:peb:slot3 speech -cart1 song.rpk

Related tools