Repository navigation

MOSS TTS integration with nearly full feature parity (MONOLITH PR WILL NOT MERGE - for reference only - use PRs 501-507 instead) - #500

Closed

voxmenthe wants to merge 33 commits into

Blaizzy:mainfrom

voxmenthe:moss-tts-integration-part-1

Commits on Feb 13, 2026

Add full MOSS tokenizer codec and streaming decode + some scaffolding
voxmenthe
committed
634fcf2
fix codec decode issues and clean up scaffold + misc
voxmenthe
committed
df330b5
fix decode layout normalization and replace tests
voxmenthe
committed
6a8742d
Added full runtime modules for text to backbone to local transformer to audio tokens to codec decode to audio
voxmenthe
committed
867689f
Fixes and runtime scaffolding
voxmenthe
committed
faf32e0
Code fixes to unblock real-model test runs
voxmenthe
committed
65ad946

Commits on Feb 14, 2026

update model hf
voxmenthe
committed
1a2dc73
code and test stabilization
voxmenthe
committed
1a05642
Stabilization for stream sinc contract and waveform normalization
voxmenthe
committed
eb36d46
add duration and local VQ inference controls for Moss TTS
voxmenthe
committed
e011bb3
add fixtures
voxmenthe
committed
c1adc0c
add moss fixtures
voxmenthe
committed
a15d24c
fixed decode-layout tie resolution
voxmenthe
committed
7407487
Implemented new isolated runtime package
voxmenthe
committed
a18c294
add runtime gate coverage and and fix santize remaps and add realtime compatibility and fix codec square tie orientation
voxmenthe
committed
99b2795
fix decode/crossfade bugs
voxmenthe
committed
f7059ce

Navigation Menu

Repository navigation

MOSS TTS integration with nearly full feature parity (MONOLITH PR WILL NOT MERGE - for reference only - use PRs 501-507 instead) - #500

MOSS TTS integration with nearly full feature parity (MONOLITH PR WILL NOT MERGE - for reference only - use PRs 501-507 instead)#500
voxmenthe wants to merge 33 commits into
Blaizzy:mainfrom
voxmenthe:moss-tts-integration-part-1

Commits on Feb 13, 2026

Add full MOSS tokenizer codec and streaming decode + some scaffolding

fix codec decode issues and clean up scaffold + misc

fix decode layout normalization and replace tests

Added full runtime modules for text to backbone to local transformer to audio tokens to codec decode to audio

Fixes and runtime scaffolding

Code fixes to unblock real-model test runs

Commits on Feb 14, 2026

update model hf

code and test stabilization

Stabilization for stream sinc contract and waveform normalization

add duration and local VQ inference controls for Moss TTS

add fixtures

add moss fixtures

fixed decode-layout tie resolution

Implemented new isolated runtime package

add runtime gate coverage and and fix santize remaps and add realtime compatibility and fix codec square tie orientation

fix decode/crossfade bugs

Commits on Feb 15, 2026

add segmented long-form generation with bounded continuity and test coverage

Enable segmented long-form generation to stream chunk results while preserving

harden codec tie handling and realtime load/prefill contracts

add MOSS preset registry and complete CLI/server promotion for missing fields (preset, input_type, quality, sound_event, ambient_sound, language)

add multilingual evidence tests and readme

restore true incremental streaming for moss_tts_realtime and fix square token-layout ambiguity and add examples

update moss documentation and add convert script

a few fixes and tests

align realtime prompt/sampling behavior and document new session controls

Implemented input/pronunciation semantics end-to-end and updated docs.

update docs and add moss showcase examples

TTSD continuation and realtime session fixes

harden realtime stream assembly and speech kwargs validation

apply black and isort formatting

chore: update pre-commit hooks

black isort formatting

add deterministic generation recipes and seedable CLI controls