16 β Development Timeline
Reconstructed from git history (git log --all). Dates use the commit
author timezone. Phase headings follow the βPhase Nβ tags used in the
commit messages and initialplan.md.
Phase 0 β Project Foundation
| Date | Commit | Summary |
|---|
| 2025-10-25 | e2fa4cc | Initial commit. |
| 2026-02-09 | 18f1c93 | Phase 0: directory skeleton, config template, requirements files, setup scripts. |
| 2026-02-09 | 6b9d003 | Merge remote, resolve .gitignore conflict. |
| 2026-02-09 | d33bce8 | Consolidated to two venvs: venv_main and venv_tts. |
Phase 1 β Core Infrastructure
| Date | Commit | Summary |
|---|
| 2026-02-09 | 8bb8326 | Phase 1: config loader, SystemState, event dispatcher, latency tracker, GPU monitor, logging. |
| 2026-02-09 | 1461248 | Fix config path resolution; remove plan files from tracking. |
Phase 2 β UI Shell
| Date | Commit | Summary |
|---|
| 2026-02-09 | 85a7a75 | PySide6 main window, A/B sections, custom widgets, dark theme. |
Phase 3 β STT Module
| Date | Commit | Summary |
|---|
| 2026-02-09 | 3450a3e | FasterWhisper engine + worker. |
| 2026-02-09 | ec3d850 | Hallucination filter list and gaming phonetic index. |
| 2026-02-09 | 971bf66 | Live (30 %) + accumulated (70 %) text panel split. |
| 2026-02-09 | 65418b0 | STT term-file drag-drop, initial_prompt, max_segment_age, VAD log suppression. |
| 2026-02-09 | d6116c5 | Adjacent dedup + short-phrase repeat filter to fix repetition. |
Phase 4 β LLM Translation
| Date | Commit | Summary |
|---|
| 2026-02-09 | 1a42dd5 | Engine, translator, controller, UI tab. |
| 2026-02-09 | 3c425cd | CUDA libs, ChatML prompt, manual-input mode. |
| 2026-02-09 | 5a9fbb8 | OpenAI-compatible API backend. |
| 2026-02-09 | 6a45a0c | Term capacity control + status display in STT tab. |
| 2026-02-09 | 6b1ff45 | Vertical scrollbar in settings tabs. |
| 2026-02-09 | 46675dcβ2e3bcef | Combobox arrow / button visibility polishing pass. |
| 2026-02-09 | d0d5577 | Plan update: 4.1, 4.2, 4.5, 4.6 done; 4.3, 4.4 deferred. |
| 2026-02-10 | d34045f | Phase 4.4: RAG vector retrieval for large glossaries. |
Phase 5 β TTS Integration
| Date | Commit | Summary |
|---|
| 2026-02-09 | 6fe4937 | Phase 5: GPT-SoVITS module skeleton. |
| 2026-02-09 | 7df8322 | Suppress pynvml deprecation noise. |
| 2026-02-09 | 857f3fb | Fix SoVITS startup: correct PYTHONPATH and project root. |
| 2026-02-09 | 5e429ee | Wire TTS to translation events; default reference audio. |
| 2026-02-09 | 4a19037 | Use absolute ref_audio_path; better error logging. |
| 2026-02-09 | 0f22fcc, 142f915 | TTS warm-up on server start, moved to a background thread. |
| 2026-02-09 | 803bf2c | Fix STT repetition, audio device selection, TTS latency display. |
| 2026-02-09 | facf711 | Phase 5 complete: end-to-end STT β LLM β TTS working. |
| 2026-02-09 | f688df7 | TTS streaming playback + 25 ms latency target. |
| 2026-02-09 | 74be837 | Audio device race fix, dedup, real TTFA measurement. |
| 2026-02-09 | e9d4556 | Phase 5 overhaul: two-tier playback. |
| 2026-02-10 | 53e3cbe | Raise max_segment_age from 2 s to 4 s. |
Phase 6 β Pipeline Orchestration
| Date | Commit | Summary |
|---|
| 2026-02-10 | f5febe2 | Pipeline coordinator + file output (TXT / SRT / WAV). |
Phase 7 β OBS Overlay
| Date | Commit | Summary |
|---|
| 2026-02-10 | c79f52c | OBS Browser Source overlay server. |
| 2026-02-10 | 27f7da7 | Stroke / shadow controls; standalone subtitle timing. |
Phase 8 β Launch and Training Tab
| Date | Commit | Summary |
|---|
| 2026-02-10 | 2d1ee60 | Launch scripts, graceful shutdown, training tab placeholder. |
Phase 9 β Quality Pass
| Date | Commit | Summary |
|---|
| 2026-02-10 | 065d7cf | TTS queue pruning, translation context, glossary corrector, benchmarks. |
| 2026-02-10 | 049f1ab | KB tab consolidation; SRT-based training tab. |
| 2026-02-10 | eb263e5 | Refactor KB tab: drop Quick Add; parallel file table + STT term editor. |
| 2026-02-10 | 554d4be | Fix delete-button width in KB tab. |
| 2026-02-25 | 890b782 | Hallucination filter UI editor; fix thread shutdown races. |
Diarisation and Qwen3-TTS
| Date | Commit | Summary |
|---|
| 2026-03-11 | 8a10303 | Speaker diarisation + Qwen3-TTS integration. |
| 2026-03-11 | 96711b3 | Connect translation target language to TTS. |
| 2026-03-25 | 4f30868 | Qwen3-TTS modes and status handling. |
Simple / Pro Modes
| Date | Commit | Summary |
|---|
| 2026-03-29 | b610cbc | Mode toggle and Simple control panel. |
| 2026-03-29 | 0a9cc55 | Staged startup, progress states, rollback. |
| 2026-03-29 | 5ca6555 | Status pills, keep-models-loaded stop option. |
| 2026-03-29 | 6db6b27 | Tooltips and validation checklist. |
| 2026-03-29 | 9b8f623 | Tab text panes; metrics in status area. |
| 2026-03-29 | 2bee0c9 | Compact quick-settings; single-line metrics. |
| 2026-03-29 | e96bcb9 | Separate Pro / Simple Section A layouts. |
| 2026-03-29 | 81cedf4 | Stage timeouts; extended Qwen3 warm-up. |
| 2026-03-29 | fdf978c | Force Qwen3 custom-voice preset for Start All. |
| 2026-03-29 | 2311042 | Reserve stdout for protocol; drain stderr. |
| 2026-03-29 | ff46a36 | Qwen3 synth + playback debug logs. |
| 2026-03-29 | 123b65e | Tune Qwen3 preset for lower synth latency. |
Echo Cancellation
| Date | Commit | Summary |
|---|
| 2026-04-02 | d4055ca | Clear-panel baseline reset for source / translation. |
| 2026-04-02 | f334773 | Playback timeline, NLMS adaptive filter, double-talk detection. |
At a Glance
- 68 commits across the rewrite, starting from the initial commit
on 2025-10-25 and the Phase 0 scaffolding on 2026-02-09.
- The end-to-end happy path was working by 2026-02-09 (
facf711); the
remaining months were quality, robustness, additional backends, and
UX modes.