Yesterday
-
card: note that the windows server build carries the 0.1.2 fixes
-
server: rebuild windows engine with the 0.1.2 fixes (second-render mask, caching, voice paths, skip-first-step, drop CFG skip)
-
card: document patch 0.1.2
-
patch: re-release 0.1.2 (second-render bug, reference/text caching, voice switching, GUI voice list, skip-first-step, drop CFG skip)
-
patch: 0.1.2 re-release - fixed ggml DLLs, local-attention switch removed, CFG truncation + uncond reuse only
-
patch: 0.1.2 rebuild - fixed ggml DLLs (stale vocoder conv kernel caused DC offset / electric timbre) + voices manifests
-
Upload GSV.CPP-win64-patch-0.1.2.zip
2 days ago
-
docs: language values / mixed-script routing / non-ASCII paths
-
server: build b36f6dc (non-ASCII paths, mixed-language frontend)
-
README: patch 0.1.1 note
-
patch 0.1.1: mixed-language modes, non-ASCII paths, auto_yue
3 days ago
-
licenses: GAG (MIT)
-
notices: GAG (MIT)
-
README: credit GAG (MIT) and the upstream projects
-
windows one-click package (GAG + engine + embedded python + models)
-
README: link build docs
-
docs: api/benchmark/usage
-
licenses: G2PW Apache-2.0
-
notices: G2PW (Apache-2.0)
-
README: server/frontend/windows sections, zip hash
-
docs: api/benchmark/usage
-
prebuilt gsv_server (windows/linux, vulkan)
-
frontend data: g2pw, bert tokenizer, nltk, fast-langdetect
4 days ago
-
Update README.md
-
vocoder: ggml sub-pixel GGUF (v0.1.2) replaces ONNX; refresh README/docs
-
vocoder: remove the ONNX model (replaced by gsv-vocoder-f32.gguf)
5 days ago
-
Upload 18 files
-
initial commit
