Notable changes per release. Full history lives in the commit log.
3.0
New
AI providers and model picker. OpenRouter, Gemini, OpenAI, xAI, Anthropic, DeepSeek, Groq, Ollama, a local proxy or any OpenAI-compatible endpoint. Primary and backup order, Fast / Smart / Vision models per provider, keys encrypted with Windows DPAPI.
Pro style tab. Real font previews, colors, outlines, shadows and 3D, glow, 10 background shapes, active word, position and tilt. Quick effects and reset per group.
Saved styles and last-used style. Save, rename, duplicate, favorite, export and import styles. New videos start from the last style used, remembered separately for vertical and horizontal.
51 templates with a moving gallery. Grouped templates that animate on hover, plus random and mix styles.
Copy style from an image. Paste a caption screenshot and get matching colors, a similar font, case, position, highlight and animation.
Brand kit with brand colors, fonts, brand words, emoji and tone, plus AI suggestions for words to highlight.
Animation studio. 40 ready animations (in, active word, out, loop) with moving previews and a keyframe editor.
Sound effects per word. Automatic sounds on key words, numbers, emoji or every page. 12 built-in sounds or your own, on one SFX track.
Fixed
Saved styles and the last-used style can no longer be lost when two processes write at the same time.
The saved-style card menu stays open while the list refreshes in the background.
An empty saved-style list no longer hangs on "loading".
Known limits
Heavy templates (3D, long shadows) on very long videos need extra preparation time.
Templates designed for 16:9 look small on 9:16. Use the "fits this video" filter.
Copy style picks a similar font from the bundled collection, not the exact font.
2.0
New panel and engine
Rebuilt panel with a home screen grouped by task, sibling tabs, a command palette, history, and a shared review list for every tool.
Local Python engine with JSON progress, cancel, a persistent worker that keeps Whisper warm, a GPU lock and a disk-space guard.
Everything works on a clone. The original sequence and media are never changed.
Tools
Silence cutting with an automatic threshold, word guard and four presets.
Filler removal with a second listening pass and an acoustic detector.
Repeated-take removal that keeps your last take, with a listen-before-apply preview.
Profanity censor with beep, mute or duck.
Animated word-by-word captions rendered as a transparent overlay.
Auto chapters, viral clip finder, auto zoom, auto resize, virtual angles, podcast multicam and b-roll.
Clear Voice: DeepFilterNet3 denoise, click and breath removal, loudness per platform and music ducking.
Klipora