Skip to content

Releases

VideoFlow 0.4.0 — new Studio & Settings (Mac Electron preview)

Choose a tag to compare

VideoFlow 0.4.0 brings the latest Studio and Settings redesign to the downloadable Electron app.

  • Canvas + contextual Details/Chat, inline character and identity editing, visible AI session selection, and a focused mobile/browser layout.
  • Settings grouped into General, AI connections and Advanced; simpler navigation throughout the app.
  • Native Codex/Claude sessions for text planning, fal.ai image/video routing, selectable BytePlus video, saved character voices, script captions and optional HyperFrames composition workflow.
  • Fixes for stale Cancel values, wrong-scope chat actions and delayed responses after session switching. Desktop startup now finds standard user CLI install locations.

Choose arm64 for Apple Silicon Macs and x64 for Intel Macs. Quit the older app before replacing it; projects remain in the existing user data folder. Windows installers are not included.

Mac downloads are ad-hoc previews, not Apple-notarized releases. macOS may require System Settings → Privacy & Security → Open Anyway. Managed Macs may prevent that approval. The package does not disable Gatekeeper. Linux installers are not included in this publication: the native builds passed their backend/frontend tests, but runner DNS failures prevented downloading Ubuntu dependencies for the clean-install smoke test. Use the current source/browser route on Linux until installers can be verified. SHA256SUMS.txt accompanies the installers.

The app includes its current Go backend, interface, FFmpeg/FFprobe and fonts. Media binaries and their corresponding source/license bundle are reused only after checksum verification and a check that the locked media build sources are unchanged; the application backend and frontend are rebuilt from this release commit.

For AI planning, install and sign in to the official Codex or Claude Code CLI on the computer running the app. These optional account tools are not bundled. Select/create a session in Studio. Add media provider credentials under Settings → AI connections. Paid generation still requires explicit Run/review controls; automated release checks do not spend generation credits. HyperFrames additionally needs its documented optional runtime.

Verification: the Mac artifacts were built from commit 9bb49ce in workflow run 43058107849. Both Mac jobs passed, including extracted ZIP signature and native-window/shutdown checks. The shared workflow was cancelled after Linux installation was blocked by Ubuntu archive DNS failures; this is a deliberately Mac-only publication, not a claim that the full matrix passed.

VideoFlow v0.3.2-browser-preview.1 — fully bundled

Choose a tag to compare

VideoFlow browser studio — repaired workflow preview

These are native browser-launcher ZIP packages for macOS and Linux, not Electron DMG/DEB installers. Choose darwin-arm64 for Apple Silicon, darwin-amd64 for Intel Mac, linux-amd64 for x86 Linux, or linux-arm64 for ARM Linux. Extract the entire archive and open Start VideoFlow; keep the launcher terminal open. The app opens at http://127.0.0.1:8000.

Each package includes the Go app, browser interface, FFmpeg, FFprobe, caption fonts, START HERE.txt and llms.txt. No Go, Node.js, Python or FFmpeg installation is needed. The standalone llms.txt and SHA256SUMS.txt are also attached to the release.

This preview contains the repaired character/reference upload and generation flow, canonical runnable chat actions, truthful readiness and model preflight, durable paid-request replay/recovery, real caption exports and ordered clip assembly. Old downloaded packages do not automatically receive these changes.

The preview is not Apple-notarized. macOS can block its first launch; use the system Open Anyway option only for a download you trust. Native Electron installers require a separate signing/release workflow. Windows native packages are not provided.

Start by testing project creation, reference upload and persistence without submitting generation. Add your OpenRouter key in Settings only when ready for a small budgeted test. Internet access and OpenRouter credits are required for AI generation and transcription. Live model/device testing is not part of the automated packaging checks. If a submission is uncertain, review saved work and provider activity before retrying.

Projects, media and your saved API key stay in the private user data folder across updates. Back it up. Do not run multiple app copies against that folder.

All four archives must pass native build/tests and the extracted-app smoke check before publication. Checks verify UI/API startup with an empty system PATH, bundled media tools, audio extraction and caption rendering. Each archive includes license texts and matching media sources/build scripts in media-sources.zip. FFmpeg is a separate process built with GPLv3 components.

VideoFlow 0.3.1 — macOS ad-hoc preview (manual approval required)

Choose a tag to compare

This rebuild fixes the invalid/incomplete macOS bundle signatures in version 0.3.0. Both the Apple Silicon and Intel apps, their frameworks, and the bundled Go/FFmpeg/FFprobe executables are signed consistently with ad-hoc signatures. The final ZIPs are extracted and checked with strict recursive signature verification on their native Mac runners before publication.

This is an ad-hoc preview, not a Developer ID-signed or Apple-notarized release. macOS still needs your explicit approval; the package does not disable Gatekeeper. A company-managed Mac can prohibit that approval.

  1. Download the DMG matching your Mac: arm64 for Apple Silicon, x64 for Intel.
  2. Quit the older VideoFlow app and replace it in Applications with this new version. Your user data folder remains separate.
  3. Try opening VideoFlow. If blocked, use System Settings → Privacy & Security → Open Anyway and confirm. Apple's instructions.
  4. Add your OpenRouter key in VideoFlow Settings.

The DMG includes a first-launch guide. Go, Node, Python and FFmpeg do not need to be installed separately. The interface, engine and media bundle are unchanged from the tested 0.3.0 release; this rebuild corrects packaging/signatures. Linux and web deployment remain available through the 0.3.0 release and repository. Default Gatekeeper acceptance still requires an Apple Developer ID certificate and notarization.

VideoFlow 0.3.0 — Electron desktop and web

Choose a tag to compare

@github-actions github-actions released this 02 Oct 05:01

Known macOS installation issue: the macOS packages in this preview have invalid or missing code signatures and are not notarized. macOS may report them as damaged. Use the rebuilt macOS 0.3.1 ad-hoc preview, which repairs the bundle signatures and includes first-launch instructions. That preview still requires manual macOS approval; normal Gatekeeper trust requires Developer ID signing and notarization. Linux packages are unaffected.

VideoFlow now has a desktop app window powered by Electron, sharing the Svelte interface and Go backend with the web version.

Download the package matching your computer:

  • macOS: DMG or ZIP, arm64 for Apple Silicon; x64 for Intel.
  • Ubuntu/Debian Linux: DEB installer, arm64 or x64. This installer configures the Chromium sandbox and AppArmor integration.
  • Other Linux distributions: AppImage or tar.gz. These need standard graphical desktop libraries and a distribution policy permitting Chromium user namespaces; AppImage can also require FUSE.

The app includes its Go backend, frontend, FFmpeg, FFprobe and caption fonts. You do not need to install Node, Go, Python or FFmpeg. Open Settings and add an OpenRouter API key with credits to use AI generation; AI requests require internet access.

Projects and settings stay on your computer in the existing VideoFlow user configuration directory. Closing the app shuts down its engine. Avoid running the older browser launcher and Electron against the same data directory simultaneously.

macOS packages are unsigned unless signing credentials are configured, so macOS may request explicit approval before first use. Each OS/CPU package is built on a native runner and tested for app rendering, backend cleanup and bundled media processing. SHA256SUMS.txt contains download checksums.

Developers can also run the separate web frontend/backend containers using docker compose up --build -d, then open http://127.0.0.1:8080. Remote deployments need an authenticated HTTPS proxy. This release does not deploy a remote server.

VideoFlow v0.2.0-openrouter.2 — fully bundled

Choose a tag to compare

VideoFlow now bundles the app, browser interface, FFmpeg, FFprobe and a caption font in each macOS/Linux download.

Download the ZIP for your computer, extract the entire folder and open Start VideoFlow. Enter your OpenRouter key in Settings. No Go, Node.js, Python or FFmpeg installation is required. Projects and your saved key stay in your user data folder when you update.

Internet access and OpenRouter credits are required; AI generation is billable. This is an unsigned preview, so macOS may show a security prompt.

Each archive includes third-party license texts and the exact media source archives/build script in media-sources.zip. FFmpeg runs as a separate process and is built with GPLv3 components. The source code and build checksums are available in the same download.

All four downloads are built and checked on native macOS/Linux runners. Tests launch the extracted app without system tools on PATH, encode video, extract audio and render captions with the included font. Paid live generation is not part of these packaging checks.

VideoFlow local preview — OpenRouter

Pre-release

Choose a tag to compare

@JienWeng JienWeng released this 01 Oct 14:53

VideoFlow now routes AI generation through OpenRouter and includes a browser-based desktop download.

Download the ZIP for your operating system, extract the whole folder, then open Start VideoFlow. Enter your OpenRouter key in Settings. No Go or Node installation is needed. Projects remain in your user data folder when you update.

This is an unsigned preview. OpenRouter credits are required and generation is billable. FFmpeg is not bundled; install it for transcription, frame QA and caption burning. See START HERE.txt inside the download for details.

Automated provider/workflow tests and a live authentication/catalog check passed. Paid generation has not been verified end to end against the live account.