Skip to content
Snippets Groups Projects
Select Git revision
  • 1.0.1
  • 1.0.x
  • 1.1.1
  • 1.1.2
  • 1.1.x
  • 1.4.x
  • analysis_fix1
  • analysis_fix2
  • burg_optimization
  • cng_fix
  • cng_fix2
  • draft
  • draft_wip
  • dred_bitrate
  • dred_bitrate2
  • dred_bitrate3
  • exp-fwgan-plc
  • exp-fwgan1
  • exp-fwgan10
  • exp-fwgan2
  • v1.5.2
  • v1.5.1
  • v1.5
  • v1.4
  • v1.3.1
  • v1.3
  • v1.3-rc2
  • v1.3-rc
  • v1.3-beta
  • v1.2.1
  • v1.2
  • v1.2-rc1
  • v1.2-beta
  • v1.1.5
  • v1.1.4
  • v1.2-alpha2
  • v1.2-alpha
  • v1.1.3
  • v1.1.2
  • v1.1.1
40 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.022Sep2120161513125111Aug9843131Jul302827262423222012111097330Jun2827232221201918161514131265131May30292827262524232220191817161514131211109827Apr262019171422Mar17107621Feb1531Jan272625201713121121Dec2019161398762130Nov28242310987431Oct262524212019181454230Sep2928272416157654Aug24Jul221512109765bugfixpulse-related bugfixModel updateexp_densenet3exp_densenet3Using a DenseNet for DREDmore lavoce stuffHandle the case where the initial state didn't fitupdated modelQuantizing initial state with rdovae tooadded FARGAN adversarial training scriptadded FARGAN adversarial training scriptexp_fargan33exp_fargan33Bunch of updatesupdated modelexp_rdovae_state2exp_rdovae_state2Handle the case where the initial state didn't fitReplace pvq with rdovae for initial statequantizing initial state with rdovae tooBand-limiting the voicing parameterremoved trailing whitespace in farganintegrated JM's FFT ada convRemove --has-gain and --passthrough-size argsAdd conditioning interpolation, fwconv layerOnly use one frame of pre-loadingRandomly double the training sequence lengthAdds skip connections5-tap pitch predictorAlso use previous frameexplicit signal gain, explicit pitch predictorFARGAN initial commit in Opusfixadded requirements.txt to osceadded more enhancement stuffadded copyright headersadded LPCNet torch implementationbrought NoLACE up to dateFixes vnni macro redefinition with clangrenamed ShapeNet to NoLACEFixes stack overflow for some custom modesChanging DRED exp. ID so we can reserve 127Don't allocate DRED bits if we don't have enoughMaking it easier to remove DRED experimental IDimproved auto-scaling in wexchange
Loading