A video that knows who you are | v0.7.2 (all updates since 0.6.0)

Mousify v0.7.2 — A video that knows who you are

Build the character first. Then the narration.

This post covers everything since 0.6.0 — versions 0.7.0, 0.7.1 and 0.7.2 together.

Same screen recording — why does someone else's commentary sound like a person, and yours like a manual? The difference isn't the edit. It's who is talking. 0.7 turns that into a feature.

🎭 Try this first: the character roster

The narration stops sounding like "the AI" and starts sounding like the person you defined.

• Several characters, switch anytime: each keeps its own persona, and projects remember who spoke last
• 11 archetype templates: gaming, knowledge and general sets apply in one click — then make them yours
• Speech flavor and zodiac: regional wording, favorite phrases — fill them in and the voice genuinely changes
• Visible provenance: every generated line shows which persona detail it drew on

⚔️ Game commentary that reads the opponent, not just you

World of Warcraft battleground commentary gets a major upgrade. It reads the game's own combat log — nothing is guessed from pixels.

• Reactions to the enemy: burst cooldowns popped on you, Sprint or Blink escapes, crowd control landing — the narrator comments like a player, and admits it when you forgot to dodge
• Flags are the storyline: pickup, drop, hand-off to a teammate, killed while carrying — four distinct situations
• Emotion on the edge: deaths are always told; clutch heals, praising teammates, surviving by a hair
• Density that fills the pace: long empty stretches fast-forward instead of padding

☁️ Cloud voices: MiniMax, Doubao — and your own

• Two new providers: MiniMax and Doubao (Volcano Engine), bring your own key; Doubao 2.0 is the best-sounding Chinese voice we've heard
• Dialect voices: Northeastern, Cantonese, Sichuan, Beijing — to pair with the persona's speech flavor
• Cloned voices: clone your voice on Volcano, paste the voice ID, and narrate your own videos in your own voice
• Live preview and progress: cloud voices preview on click; batch voicing shows per-line progress and an ETA

📝 Long videos draft properly — and fit

• Windowed drafting: an hour-long tutorial is drafted in ~5-minute windows and stitched for continuity — no more "only 12 lines"
• Fit to time: lines that don't fit are shortened in batches by AI and re-voiced automatically; leftovers are reported, not hidden
• Smooth zooming: 10-minute-plus timelines no longer freeze

🌐 Subtitle translation with official terminology

• Official localized terms: spell names and menu items are no longer translated literally
• Never longer than the source: a translation may not take longer to read than the original line — words are cut before the caption overflows

⚡ Faster export, with progress all the way

• 720p 2.6× faster, 4K 1.4× faster: smart zoom renders at delivery resolution and encoder settings converge; muxing 40 s → 4.5 s
• Multi-part exports report every part
• Game projects skip auto spotlights: the cursor isn't the point in a game

✨ Text and picture

• Caption backgrounds: solid, gradient, rounded, directional, with independent opacity; text itself can use gradient fills
• Full-frame spotlight inside insert cards; the playhead rests reliably on card boundaries

🩹 Stability (0.7.2)

• Blank settings window heals itself: reopening the settings window from the tray could occasionally show an empty window; if the page has not loaded within 8 seconds it reloads, and if that fails it is rebuilt once
• Tray "Quit" always quits: in a rare stuck-window state Quit did nothing; if the app has not exited within 2 seconds the process is now terminated, cursors restored as usual
• No raw backend error for a deleted scheme: the library refreshes and says so

🔧 Fixed: "I selected it, nothing happened"

• Select-all on a voice track then Delete really deletes the batch; Ctrl+Z restores it; toolbar 🗑 and keyboard agree
• A line you added with "Add text" can be voiced on its own instead of being treated as the whole caption layer
• Fit-to-time silently doing nothing and drafts capped at 12 lines — fixed; timeline zoom freezes are gone for good, milliseconds even on 20-minute projects with silence cuts
• Insert-card preview: black frame at start, seam race, and card ending — three fixes

🎚 Editing feel: changes that came straight out of real edits

• Voice levels finally match: mic takes, cloud voices and on-camera audio are all matched to one loudness as they land on the track; older clips can be matched with one click
• A volume line on every voice clip: drag the line inside the clip up or down to set that clip's level, the way you'd drag a system volume slider
• Trimming a segment tracks your cursor: the block and its duration follow the drag instead of jumping only when you let go
• Frame-exact stickers and captions: park the playhead on the frame you want, then press ] for the out point or [ for the in point; dragging an edge snaps to the playhead too
• Every voice-clip action is one right-click away: match loudness, denoise, split at the playhead, transcribe, delete — no more hunting through panels
• Room noise can be removed: three levels for steady noise (fans, electrical hum, air conditioning); if the take is already clean it says so instead of dulling your voice
• Alt + drag = duplicate: hold Alt and drag a caption or sticker to pull off a copy that lands wherever you release it
• Ctrl + wheel zooms the timeline, anchored at the mouse position
• Alt + wheel adjusts track height: scroll over the track-name column and the audio, mic, voice and clip tracks grow or shrink together, so waveforms are easier to trim (the Premiere gesture)

Wherever it goes, it owns the stage. This time, the one talking is you.