Every release, every change, in plain English. No "synergies," no "unlocking value." Just what got better and why.
Transcription is about two and a half times faster and keeps a lot more of what you actually said - three separate filters were deleting real speech. Translation no longer takes longer than the transcription it follows. Eight more local models, rebuilt settings, and brightest-frame thumbnails that are instant after the first pass.
Drag the thumbnail text where you want it, and drag the upload queue into order. Moving a video from #45 to #2 is one drag now instead of dozens of arrow clicks.
Five one-click thumbnail styles, each driving a properly layered renderer, plus a backing plate for logos that used to disappear into busy footage.
Out of a full audit. Stored credentials are now encrypted against your Windows account, YouTube sign-in got CSRF and PKCE protection, and the bug behind 'uploaded but never appeared on the channel' is fixed at the root.
A one-line patch. The help-video playlist moved to a new address, so the Dashboard button pointing at it needed updating.
A single video with no speech in it no longer takes down the rest of the batch. The Claude model list is current again. YouTube sign-in works for Brand and Workspace accounts, which it quietly didn't before.
The big one. Premiere-style EDL import. Live dark/light theme swap with user-loadable JSON themes. Caption pipeline rework with semantic chunking and per-word timing for translated languages. Transcription accuracy overhaul. GPU work moved into a crash-isolated child process. Anti-clickbait thumbnail prompts. Tab-level translator caching so batch uploads stay fast. Installer ~180 MB lighter than v1.1.x. Every Norse module statically links FFmpeg now — no more avcodec / avformat DLLs at runtime.
Removed the Localize tab. Dropped the CUDA 12 fallback path — CUDA 13 is now the only GPU runtime. The Whisper model cleanup pass means smaller installs and fewer "which model do I pick?" decisions.
Fixed an EDL import bug that wouldn't read certain Premiere exports cleanly. Across-the-board performance pass on the Timeline Editor — scrubbing, seeking, and thumbnail-strip generation are visibly faster.
Added the Finalize Videos tab. Audio mixing across multiple recorded tracks (game audio, mic, chat) finally happens inside Loki instead of needing a side trip to a DAW.
Translation quality improvements. The CJK (Chinese, Japanese, Korean) caption flow keeps per-character timing now instead of mangling segment boundaries — which other tools quietly get wrong.
Translation features land. Transcribe in any of 99+ languages, translate to 34+ target languages via NLLB-200 — all locally, no per-minute fees, no cloud upload.
YouTube upload pipeline fixes. License-key handling cleanup. Smaller but important reliability pass for the upload flow.
Security pass and path-handling fixes. Non-ASCII video filenames (Japanese, Korean, accented characters) work cleanly through the whole pipeline now.
The first public release. Transcribe, metadata, thumbnails, upload — the four tabs that became the spine of the whole product.
Every version is listed, oldest to newest. The current build is always the safer pick,
and updates are free for life once you have bought Loki Studio.
If you bought Loki Studio, every future release is yours. No subscription, no upgrade fee.
