Skip to content

Velorn v0.2.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 09 Jul 16:07
v0.2.0
e12433a

ComfyStudio is now Velorn — and v0.2.0 is the biggest release in the project's history. The headline: Velorn is now an editor your AI agent can drive. Connect Claude Code, Codex, Cursor, or any MCP client and it can inspect your timeline, review your cut shot by shot, make previewed edits, install community workflows, generate media, and export — while you stay in the director's chair.

Downloads

  • Windows Installer: standard Windows install experience for most users
  • Windows Portable: no-install Windows build for quick testing or portable use
  • Mac (Apple Silicon): for M1, M2, M3, and newer Macs
  • Mac (Intel): for older Intel-based Macs
  • Linux AppImage: portable Linux build
  • Linux deb: Debian/Ubuntu package

Ignore the auto-generated source-code archives unless you plan to build Velorn from source.

Agents (MCP) — the headline

Velorn now ships a local MCP server (Settings > Agents (MCP), one copy-paste to connect Claude Code, Claude Desktop, Codex, or Cursor). Your agent gets 100+ tools over the open project:

  • Review an edit like a person would: read the timeline, inspect clips, capture composed frames, sample ranges as contact sheets, page through visible shots in a fast cut, check media health and export readiness.
  • Make safe editorial changes: move/trim/split/delete clips, tracks, markers, transitions, retimes, audio gain and fades — every write tool previews its plan first and applies only after approval, on Velorn's normal undo stack.
  • Build graphics: text, shapes, solids, adjustment layers, GLSL effects, and keyframes, all agent-drivable.
  • Generate and deliver: queue image/video/music generation, transcribe and render captions, run H.264 delivery exports and social delivery batches, write FCPXML.

Bring any ComfyUI workflow — your agent sets it up

Hand your agent a community workflow (a comfy.org share link, a .json file, or pasted JSON) and it can now:

  1. Analyze it — which custom nodes and models it needs, what's already on your machine, what's missing and how big the downloads are.
  2. Install the missing pieces after you approve — node packs and models, with live progress, never overwriting existing files.
  3. Run it on your timeline — your clips and assets as inputs, results imported straight back into the project.

Official ComfyUI templates get the same treatment from the Generate tab: browse the live template catalog, one-click import, in-context dependency setup, and imported workflows now appear right in Generate > Featured next to the built-ins.

Motion graphics suite

  • Track mattes (alpha/luma, normal and inverted) using the layer above as the matte
  • Speed ramps with keyframable speed on video clips
  • Motion paths with smooth (Catmull-Rom) interpolation and a draggable path overlay in the preview
  • Bezier easing editor and a value graph editor in the dope sheet
  • Corner pin for perspective placement (screens, signs, inserts)
  • Keyframe copy/paste, frame-accurate keyframe snapping, and selection-preserving undo/redo
  • Free clip speed control on a log slider (no more 0.25x steps), plus five new music-video transitions

A GPU pipeline under everything

The preview and the export renderer are now fully GPU-composited (WebGL2):

  • All managed effects, tonal grades, masks, GLSL effects, and velocity blur run natively on the GPU in both preview and export
  • Physically correct motion blur (centered shutter)
  • ~3.3x faster exports via sequential WebCodecs decode, plus Apple VideoToolbox hardware encoding on macOS
  • Smoother scrubbing with no black flashes at cuts
  • Flame-style render-in-place per clip and Render In-Out range flattening, with automatic stale-render detection

Music and audio

  • Music generation (ACE-Step) right from the timeline: style tags, optional lyrics, exact duration, BPM and key — up to four takes per run
  • Audio analysis: beat detection to timeline markers ("cut to the beat"), BPM estimation, loudness measurement before delivery
  • True stereo waveforms with per-channel lanes

Interchange

  • FCPXML import and export — move timelines between Velorn and Resolve, Final Cut, or Premiere in both directions

Send Feedback

Settings > Send Feedback: report a bug or share an idea from inside the app — no account, no GitHub required. It lands directly with the team, with an optional (and fully visible) system-info block that makes bugs dramatically easier to fix.

Notes

  • Velorn is now licensed under GPLv3.
  • Velorn still depends on a separate local ComfyUI installation for local generation; cloud workflows need a partner API key. The in-app setup wizard and your MCP agent can both walk you through ComfyUI setup.
  • Some community workflows may still require manual model setup when no download URL is available — the dependency report will tell you exactly which files.
  • MCP write actions default to preview-first; agents are instructed to get your approval before installing, generating, or exporting.
Assets 8
Loading