From 4997edecb0e0ecfe08cf7d009b6da7da66da4d8b Mon Sep 17 00:00:00 2001 From: DemonGatanjieu <207458273+DemonGatanjieu@users.noreply.github.com> Date: Sat, 17 Oct 2026 12:40:09 +0800 Subject: [PATCH] CHANGELOG: v3 and v4 voices; v5 left out until tried with a real model Co-Authored-By: Claude Opus 5.5 --- CHANGELOG.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 4ab72d2..77041f7 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -12,8 +12,8 @@ - **A combo's models are cover cards**, as in the model library: the cover on top, the name below. In a combo saved from picked nodes, the slot (Main model, LoRA…) is written on the cover and a click on the card changes the model; in a default combo, the main models and LoRAs to pick from are cover cards side by side. They used to be rows as wide as the window, with **Change** far from the name (模型卡片不再横向拉满整个窗口). ### 🎙️ Character voices (角色语音) -- **Each character card shows its GPT-SoVITS model version** (v2, v2ProPlus, v4, v5dev…) after the name, so you can tell at a glance which model a voice runs on (角色卡片显示模型版本). Needs Anomalous_TTS with v3 / v4 / v5 support (its interface 15); with an older one the cards look as before. -- **v3, v4 and v5 models can be imported**: the import no longer refuses them. Each of these versions needs its own official base model and vocoder (about 0.8 GB each, v3 about 1.2 GB); **GPT-SoVITS Settings** lists them only when one of your characters uses that version, and the import card says what is still missing for the model you dropped in. A file whose version cannot be read is refused with a clearer reason; a v3 / v4 file with an older Anomalous_TTS asks you to update it. +- **Each character card shows its GPT-SoVITS model version** (v2, v2ProPlus, v4…) after the name, so you can tell at a glance which model a voice runs on (角色卡片显示模型版本). Needs Anomalous_TTS with v3 / v4 support (its interface 15); with an older one the cards look as before. +- **v3 and v4 models can be imported**: the import no longer refuses them. Each of these versions needs its own official base model and vocoder (v4 about 0.8 GB, v3 about 1.2 GB); **GPT-SoVITS Settings** lists them only when one of your characters uses that version, and the import card says what is still missing for the model you dropped in. A file whose version cannot be read is refused with a clearer reason; a v3 / v4 file with an older Anomalous_TTS asks you to update it. ### ☁️ ComfyUI on a cloud server (云服务器上使用) - **How far it goes**: fully supported on overseas cloud servers (RunPod…). On servers in mainland China (AutoDL…) only in part: browsing, offline scans and tidying work, but Civitai is mostly out of reach there, so lookups are unreliable and covers rarely download (海外服务器完整支持；国内服务器只能部分支持). Not fully tested yet: one AutoDL server so far, no overseas one. The cloud guide ends with a section for AI assistants (what is going on, the steps, how to check, what to do when it did not work, what to tell the user about safety), so a user can hand the link to their AI.