Store full IP address without anonymization

Use access_route for IP detection in HF Spaces

Add debug-headers endpoint to check proxy headers

Add ProxyFix middleware to get real client IP

Change preferences repo to channelcorp/ko-tts-arena-db

Improve UI design with modern sophisticated styling

Add ElevenLabs v3 model and change voice to Olivia

Revert "Fix database path fallback to use /tmp instead of instance/"

Fix database path fallback to use /tmp instead of instance/

Fix unterminated string literal in typecast API URL

Update Typecast TTS API to v1 endpoint (#3)

Fix fallback to prevent duplicate A/B models

Add TTS fallback: retry with different model on API error

Increase exposure for low-vote models

Remove all Hugging Face login buttons and prompts

Merge remote changes and resolve conflicts

Improve model pairing: prioritize new and low-vote models

Add Channel Hana 1.1 TTS model

Fix JSON formatting: remove trailing comma

Expand Korean TTS examples with more customer service scenarios

Update Korean TTS example prompts with customer service scenarios

Sync from GitHub repo

Improve model pairing: match models with similar vote counts

Add Humelo DIVE 나나 voice and retire 리아 voice

Humelo: DIVE API Response Field명 변경 (#2)

Add Gemini 2.5 Flash Preview TTS (API Key based)

Disable Humelo TTS (API temporarily down)

chore: Humelo 목소리 업데이트 요청 (#1)

Update tts.py

feat: Add category filters (한영혼합, 숫자혼합, 긴문장) to leaderboard

fix: Hide inactive models from leaderboard

feat: Keep only Typecast v3 (자인), deactivate 재선 and legacy models

chore: Trigger space restart for new Typecast models

feat: Re-enable Typecast v3.0

fix: Update Typecast default model to ssfm-v30

fix: Temporarily disable Typecast v3.0 (503 errors from HF Space)

feat: Add retry logic for Typecast API (3 retries with exponential backoff)

feat: Upgrade Typecast to v3.0 API (ssfm-v30), add Jaesun & Jain voices, deactivate legacy GeumHee

fix: Gemini TTS requires service account JSON (API Key not supported), disable until JSON provided

fix: Use REST API v1beta1 for Gemini TTS, disable google-wavenet/neural2

fix: Use GEMINI_TTS_API_KEY -> GOOGLE_API_KEY for Gemini TTS (client lib reads this env var)

fix: Require service account for Gemini TTS and disable legacy Typecast model

fix: Use REST API v1beta1 for Gemini TTS (supports API Key)

fix: Gemini TTS requires service account JSON (API key not supported by client lib), disable until JSON provided

fix: Use GOOGLE_API_KEY env var for Gemini TTS (simpler than service account)

feat: Enable Gemini TTS with service account JSON authentication

fix: Add ffmpeg to Docker, disable Gemini TTS (requires OAuth2), keep REST API code for future

feat: Add Gemini TTS (Aoede voice) - Google Cloud Text-to-Speech

feat: Change Typecast voice from JaeYi to GeumHee

feat: Add Typecast TTS (JaeYi voice) - https://typecast.ai/