• Joined on 2025-12-07
Koko210 pushed to master at Koko210/miku-discord 2026-01-27 19:55:18 +02:00
55fd3e0953 Cleanup. Moved prototype and testing STT/TTS to 1TB HDD
Koko210 pushed to master at Koko210/miku-discord 2026-01-27 19:53:23 +02:00
ecd14cf704 Able to now address Miku in Cyrillic, Kanji and both Kanas, incl. Japanese honorifics
641a5b83e8 Improved Evil Mode toggle to handle edge cases of the pfp and role color change. Japanese swallow model compatible (should be).
Compare 2 commits »
Koko210 pushed to master at Koko210/miku-discord 2026-01-27 19:51:12 +02:00
c0aaab0c3a Disabled KV cache offloading on llama-server and enabled Flash Attention. Performance gains in the tens.
Koko210 pushed to master at Koko210/miku-discord 2026-01-23 17:01:54 +02:00
dca58328e4 Tuned the Japanese mode system prompt and model better
Koko210 pushed to master at Koko210/miku-discord 2026-01-23 15:02:42 +02:00
fe0962118b Implemented new Japanese only text mode with WebUI toggle, utilizing a llama3.1 swallow dataset model. Next up is Japanese TTS.
Koko210 pushed to master at Koko210/miku-discord 2026-01-22 22:09:00 +02:00
eb03dfce4d refactor: Implement low-latency STT pipeline with speculative transcription
Koko210 pushed to master at Koko210/miku-discord 2026-01-20 23:06:26 +02:00
2934efba22 Implemented experimental real production ready voice chat, relegated old flow to voice debug mode. New Web UI panel for Voice Chat.
Koko210 pushed to master at Koko210/miku-discord 2026-01-19 00:29:53 +02:00
362108f4b0 Decided on Parakeet ONNX Runtime. Works pretty great. Realtime voice chat possible now. UX lacking.
Koko210 pushed to master at Koko210/miku-discord 2026-01-18 03:35:58 +02:00
0a8910fff8 Changed stt to parakeet — still experiemntal, though performance seems to be better
50e4f7a5f2 Error in llama-swap catchall implemented + webhook notifier
Compare 2 commits »
Koko210 pushed to master at Koko210/miku-discord 2026-01-17 03:15:51 +02:00
d1e6b21508 Phase 4 STT pipeline implemented — Silero VAD + faster-whisper — still not working well at all
Koko210 pushed to master at Koko210/miku-discord 2026-01-17 00:01:26 +02:00
3e59e5d2f6 Phase 3 implemented — Text LLM can now stream to the TTS pipeline with the !miku say command
Koko210 pushed to master at Koko210/miku-discord 2026-01-16 23:37:42 +02:00
9943cecdec Phase 2 implemented and tested. Added warmup to pipeline and Miku queues tokens while the pipeline is warming up
Koko210 pushed to master at Koko210/miku-discord 2026-01-16 20:39:31 +02:00
b0066f3525 Tested Phase 1, fixed text channel blocking while in voice and implemented joining and leaving VC from Phase 2
Koko210 pushed to master at Koko210/miku-discord 2026-01-16 13:02:13 +02:00
911f11ee9f Untested Phase 1 (Foundation & Resource management) of voice chat integration
Koko210 pushed to master at Koko210/soprano_to_rvc 2026-01-13 00:44:17 +02:00
1b54e4d5e2 more graceful log and shutdown on unified script
Koko210 pushed to master at Koko210/soprano_to_rvc 2026-01-13 00:21:01 +02:00
346f9ccbda unified soprano to rvc script
Koko210 pushed to master at Koko210/soprano_to_rvc 2026-01-12 23:15:01 +02:00
5eedbb80e4 headless rvc created
Koko210 pushed to master at Koko210/soprano_to_rvc 2026-01-12 22:55:41 +02:00
942ca36252 Working with GUI, auto loopback creation, soprano streaming
Koko210 created branch master in Koko210/soprano_to_rvc 2026-01-12 22:55:41 +02:00
Koko210 pushed to main at Koko210/soprano_to_rvc 2026-01-12 22:50:40 +02:00
1ee08223e3 Delete README.md