v0.3.0

Everything since v0.2.0 — a much richer app.

Highlights

• Enterprise app shell — a left sidebar to switch between Frame photos and Make a video workspaces.
• Per-photo Framed / Original — photos added on the Frame tab are framed; photos added on the Video tab use the untouched original, each toggleable per photo.
• Full video editor — per-clip motion, color look & transition; title / text cards; drag-reorder, timeline zoom, snap-to-cut, editable clip length (up to 10 min); save any frame as a PNG.
• Audio depth — per-track volume, fade in/out, auto-duck under voice, and microphone voiceover recording.
• Export heads-up — estimated size, resolution, quality, and whether your music fits the video — plus live progress with time-left and a clean Cancel.
• Clear file guidance — explains HEIC / unsupported / non-image drops (with iPhone tips) instead of failing silently.
• Open source — MIT licensed, with a full README, screenshots, and contributor docs.

Downloads

• Windows: `Frame-Setup-0.3.0.exe`
• macOS (universal): `Frame-0.3.0-universal.dmg` or `.zip`

Builds are unsigned: on Windows click More info → Run anyway; on macOS approve under System Settings → Privacy & Security after first launch. The lossless photo guarantee is unchanged and continuously tested.

v0.2.0

Formats for TikTok / YouTube / X / custom sizes · session autosave & restore · settings presets · per-clip captions · color looks · slide-up & iris transitions · installable offline PWA (web build) · new app icon. Pixel-perfect guarantee unchanged: Still + Cut + Look:None + no caption = bit-identical frames.

v0.1.0

Migrate muxer: deprecated mp4-muxer -> Mediabunny

mp4-muxer is deprecated upstream; Mediabunny (same author) supersedes
it. Like-for-like migration of the fast exporter's muxing layer:

- Muxer/ArrayBufferTarget -> Output + Mp4OutputFormat(fastStart
  "in-memory") + BufferTarget, per the official migration guide
- addVideoChunk/addAudioChunk -> EncodedVideoPacketSource /
  EncodedAudioPacketSource .add(EncodedPacket.fromEncodedChunk(...));
  adds are async, so per-track promise chains preserve decode order
  behind the sync encoder callbacks
- our own encoding logic (pickAvc codec probe, AAC bitrate ladder,
  queue backpressure) is unchanged

New e2e regression test encodes a real 1s clip through VideoEncoder +
Mediabunny and asserts a valid MP4 (ftyp) comes out — video-only so it
runs on Linux CI too (no AAC encode in Linux Chromium). Finding along
the way: Playwright's chromium-headless-shell build crashes on
VideoEncoder.encode despite isConfigSupported returning true — browser
tests now use the full chromium build (channel "chromium").

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>