Restore upstream model card and add Russian samples

README.md

@@ -4,12 +4,10 @@ language:
 - ja
 - zh
 - en
-- ru
 pipeline_tag: text-to-speech
 tags:
-- TTS
-- Text-to-Speech
-- audio
+- TTS
+- Text-to-Speech
 ---
 
 # GLaDOS TTS(Text-to-Speech) Models
@@ -18,21 +16,12 @@ tags:
   <img src="https://github.com/WarriorMama777/imgup/blob/main/img/__Repository/huggingface/AI/GLaDOS_TTS/WebAssets_heroimage_GLaDOS_02_comp001.png?raw=true" alt="GLaDOS Text-to-Speech Model Heroimage" title="GLaDOS Text-to-Speech Model Heroimage">
 </p>
 
-## Current Russian voice comparison · 2026-09-06
-
-**Native Russian Style-Bert-VITS2 remains experimental; matching the original voice and all five styles has not been accepted.**
-
-[Listen: original model and three native Russian pilots, all five styles](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/README.md). The page includes difficult examples, audio hashes and results from two independent recognizers. The new original-prior constraint improves a separate English diagnostic but does not demonstrate Russian voice recovery; it is not promoted. Native pilot weights are not included in this update.
-
-The older CosyVoice pack below is an earlier experiment. Upstream PR #2 is closed. Its automatic scores do not establish listening acceptance.
-
 ## Overview
 Introducing the text-to-speech model of GLaDOS, the beloved (and slightly insane) artificial intelligence from the "Portal" series. This repository contains two models that capture the unique personality of GLaDOS, created based on Style-Bert_VITS2 and GPT-SoVITS. These models replicate GLaDOS' distinctive voice and speech patterns.
 
 ## Features
 - **Style-Bert_VITS2 Model**: This model is based on the emotional text-to-speech model developed in the Style-Bert_VITS2 repository. It captures the vibrant emotional expressions and speaking style of GLaDOS, bringing your text to life (even though GLaDOS herself may lack emotions). This model is an English-only version, trained to replicate GLaDOS' English voice.
 - **GPT-SoVITS Model**: This is a fine-tuned model based on the GPT-SoVITS repository. With just a few minutes of training data, it fine-tunes the Zero-shot TTS capability, resulting in improved voice similarity and realism. The model supports Japanese, English, and Chinese, enabling multilingual conversations with GLaDOS.
-- **Russian CosyVoice 3 Dubbing Pack**: Twenty curated Russian clips use Silero and RuAccent for Russian content, an English GLaDOS clip for the target voice, CosyVoice 3 `inference_vc` for voice conversion, and formant-preserving duration/F0 alignment.
 
 ## Sample
 
@@ -76,73 +65,37 @@ Multilingual samples in Japanese, English, and Chinese.
 ようこそ、私の新しい被験者さん。 I am GLaDOS, and I am here to support and guide you through this research facility. 我们有各种旨在挑战和提高您的技能的测试。一緒に課題に取り組み、成長していきましょう。 I have high expectations for your abilities, and I am excited to see what you can achieve. 我期待着从现在起与您合作。
 ```
 
-### Russian CosyVoice 3 dubbing pack
-
-The Russian pack contains 20 WAV samples, a complete bilingual manifest, per-clip prosody profiles, reproducible scripts, and an ASR/audio QA report. All 20 published clips pass the duration, peak, median-F0, format, and ASR thresholds recorded in the report.
-
-Live preview and contributor fork: [Random118/GLaDOS_TTS](https://huggingface.co/Random118/GLaDOS_TTS)
-
-> **Russian voice/style revision in progress.** The first public pack used per-line
-> game recordings as speaker prompts. The replacement pipeline anchors speaker
-> identity and median F0 to this repository's official Style-Bert-VITS2
-> `NeutralStyle` output. Upstream PR #2 is closed; this earlier A/B experiment was not accepted by listening.
-
-#### Model-anchored 0234 A/B review
-
-**Official model voice — Style-Bert-VITS2 `NeutralStyle` reference**
+### Russian Style-Bert_VITS2 samples
 
-<audio controls preload="none" src="https://github.com/WarriorMama777/imgup/raw/main/img/__Repository/huggingface/AI/GLaDOS_TTS/Portal_GLaDOS_SBV2_v1_neutral_original_en_short_comp.mp3"></audio>
+The fork adds native Russian synthesis for the original GLaDOS Style-Bert_VITS2 voice. The phrase below was excluded from training; two independent recognizers transcribed all five recordings exactly. Each sample uses one of the original style vectors.
 
-**Current published Russian baseline — game-recording prompt, voice score 0.685**
+```txt
+«Я получила ваше сообщение. Очень надеюсь, что это была шутка»
+```
 
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0234.wav"></audio>
+NeutralStyle
 
-**Earlier model-anchored candidate — model F0, voice score 0.846, exact ASR; listening acceptance unconfirmed**
+<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/v3-1000-p20-Neutral.wav"></audio>
 
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/model_anchor_review/0234/model-anchored-balanced.wav"></audio>
+StandardStyle
 
-**Model-anchored candidate — natural VC F0, voice score 0.845, exact ASR**
+<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/v3-1000-p20-Standard.wav"></audio>
 
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/model_anchor_review/0234/model-anchored-natural-f0.wav"></audio>
+DeepStyle
 
-| Version | Median F0 | Cosine similarity to model voice | GigaAM WER |
-| --- | ---: | ---: | ---: |
-| Current published baseline | 146.76 Hz | 0.685 | 0.000 |
-| Model-anchored, model F0 | 154.6 Hz | **0.846** | **0.000** |
-| Model-anchored, natural F0 | 160.05 Hz | 0.845 | **0.000** |
+<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/v3-1000-p20-Deep.wav"></audio>
 
-[Detailed 0234 comparison notes](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_CosyVoice3/model_anchor_review/0234/README.md)
+LightStyle
 
-#### Russian audio examples
+<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/v3-1000-p20-Light.wav"></audio>
 
-**0000 —** «И снова здравствуйте. Добро пожаловать в компьютеризированный центр обогащения „Апертур Наука“».
+Standard_02Style
 
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0000.wav"></audio>
+<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/v3-1000-p20-Standard_02.wav"></audio>
 
-**0109 —** «Что вы делаете? Прекратите! Я... я... Мы рады, что вы прошли последнее испытание...»
+[Detailed Russian listening comparison and provenance](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_StyleBert_Pilot/voice_review_2026-09-06/README.md)
 
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0109.wav"></audio>
-
-**0234 —** «Ладно. Послушайте. Мы оба наговорили много такого, о чём ещё пожалеем...»
-
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0234.wav"></audio>
-
-**0285 —** «Вы застряли. Посмотрим, попробует ли куб помочь вам сбежать...»
-
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0285.wav"></audio>
-
-**0553 —** «Вы знаете, Кэролайн преподала мне ценный урок. Я думала, что вы мой главный враг...»
-
-<audio controls preload="none" src="https://huggingface.co/Random118/GLaDOS_TTS/resolve/main/Models/Russian_CosyVoice3/samples/0553.wav"></audio>
-
-- [Listen to or download all 20 Russian samples](https://huggingface.co/Random118/GLaDOS_TTS/tree/main/Models/Russian_CosyVoice3/samples)
-- [Russian pipeline, reproduction steps, and validation](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_CosyVoice3/README.md)
-- [Manifest with English/Russian text, controls, checksums, and QA](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_CosyVoice3/manifest.jsonl)
-- [Machine-readable QA report](https://huggingface.co/Random118/GLaDOS_TTS/blob/main/Models/Russian_CosyVoice3/qa_report.json)
-
-This is a speech-to-speech dubbing route and sample pack; the original GPT-SoVITS and Style-Bert_VITS2 checkpoints are unchanged.
-
-## Installation and Usage
+## Installation and Usage
 Detailed installation and usage guides can be found in the respective model repositories. Both models support Python environments, and the Style-Bert_VITS2 model includes an API server for integration with other applications and tools.
 
 - Style-Bert_VITS2 Model: [Repository Link](https://github.com/litagin02/Style-Bert-VITS2)
@@ -154,4 +107,4 @@ These models are distributed under the CreativeML Open RAIL-M License. The GLaDO
 ### Awesome GLaDOS Project
 
 - [davesarmoury/Bringing GLaDOS to life with Robotics and AI - YouTube](https://youtu.be/yNcKTZsHyfA?si=1WqFFPXTydZn323t)
--  [davesarmoury/GLaDOS](https://github.com/davesarmoury/GLaDOS)
+-  [davesarmoury/GLaDOS](https://github.com/davesarmoury/GLaDOS)