Gradium phonon improved #1825
xenova
wants to merge
26 commits into
main
from
model/gradium-phonon-improved
added phonon support
bb98b056
clean-up
4cbdd7ee
fixes after review
a0774322
fixes after pr
b3d090e6
simplifications
0fdab1e5
changed voice loading
a7673cef
latest fixes based on hub changes
ef9af697
last changes
d698ee6e
Implemented voice-prefix caching
b645aeff
improved caching
bf0c9b1b
fixes after PR
167a6ce6
latest fixes
8c437b6e
dropped env from tests
a628e4dc
removed phonon_config getter
b979dbde
use flat float32 voices and drop voice caching
65e4005e
drop callback_function from the Phonon pipeline
b21673f7
port PocketTTSProcessor for Phonon text preparation
86ffe5c0
use full() for Phonon float32 buffers
3d373f27
removed source link
940fd4e5
add load_voice() to resolve repo-relative Phonon voices
a9f0b6fc
keep feedback tensors on the GPU with WebGPU
a776e065
Merge branch 'main' into model/gradium-phonon
35b18928
decode each chunk in one Mimi call and move voice loading to the app
898a69f9
Fix SpeechT5 text-to-audio: load its processor and keep the 1D vocode…
7c79191b
updates
773a5548
Merge branch 'main' into model/gradium-phonon-improved
ce3cd146
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub