Automatic subtitle generation (PocketSphinx) - lmmx/devnotes GitHub Wiki
- Download swig -
apt-get install swigfor ease - Download Sphinxbase (currently v.5, check
sourceforgeGitHub for latest)mv sphinxbase-XXXX sphinxbase; cd sphinxbase./configuremake
- Now its dependency is satisfied, add PocketSphinx via GitHub
-
run
./autogen.sh -
./configure -
make clean all../../src/libpocketsphinx/.libs/libpocketsphinx.so: undefined reference to `cmd_ln_print_values_r' -
make check -
sudo make install
-
- Install
maven2and then build sphinx4 - Make a test video clip, transcribe it manually, convert video to proper format:
ffmpeg -i Short_clip.mp4 -acodec pcm_s16le -ac 1 -ar 16000 Short_clip.wav
-
Use long audio alignment being developed in Sphinx
-
videogrep --input Short_clip.mp4 --transcribeproducesShort_clip.mp4.transcription.txt -
Manually transcribe to
Short_clip_manual_transcription.txt
java -cp sphinx4-samples/target/sphinx4-samples-1.0-SNAPSHOT-jar-with-dependencies.jar edu.cmu.sphinx.demo.aligner.AlignerDemo Short_clip.wav Short_clip_manual_transcription.txt en-us-generic cmudict-5prealpha.dict cmudict-5prealpha.fst.ser