Drop photos. Pick a vibe. Your video cuts to the beat—automatically. No timeline, no editing, just AI-powered cinema on Apple Silicon.
Most apps place images at fixed intervals. SteadyFlow listens to your music and places transitions on actual beats, downbeats, and song segments. 15 photos over a 2-minute song? The algorithm selects the 15 strongest musical moments automatically.
Every single image is analyzed: saliency detection finds the subject, horizon detection ensures level framing, aesthetic scoring identifies your best shots, and subject isolation creates depth masks for parallax effects.
No timeline. No keyframes. No manual adjustments. Drop your photos, pick a style card, and export. The AI Director handles pacing, transitions, color grading, and motion—all tuned to the emotional arc of your chosen music.
Everything runs 100% on-device. Your photos are analyzed by Apple's Vision framework and CoreML—never uploaded anywhere. No cloud. No subscriptions. No data leaves your Mac.
On macOS Tahoe and later, SteadyFlow harnesses Apple Foundation Models running entirely on your device. The on-device LLM analyzes your photo collection and intelligently directs the cinematic style—choosing pacing, transitions, and emotional tone based on what it sees in your images.
Forget timelines and keyframes. SteadyFlow uses high-performance GPU acceleration for real-time color grading, depth synthesis, and motion blur rendering. The CPU is reserved strictly for the final video stitching, ensuring a fluid, edit-free experience that generates professional results automatically.
On macOS 26+, Apple Foundation Models analyze your photos and craft a cinematic vision—determining style, pacing, and emotional arc. The on-device LLM sees labels, aesthetics, and composition to direct your video like a real cinematographer. Efficient heuristic fallbacks ensure great results on earlier macOS versions too.
SteadyFlow doesn't just "detect beats." It intelligently switches between Energy Analysis (Peak-to-Average Power Ratio > 4.5 for drums) and Spectral Flux (for acoustic/classical) to ensure your cuts hit the musical transients exactly where they feel right.
By extracting embedded depth maps from HEIC files or estimating them via ML, the engine separates foreground from background. This enables true 3D Parallax effects, adding a cinematic "dolly" motion to otherwise static images.
Instead of static overlays, the color engine generates 3D Color Cubes procedurally directly in memory. Whether it's the teal-shadow push of "Teal & Orange" or the warmth of "Kodak Gold," grading is applied mathematically at the pixel level on the GPU.
Your photos never leave your device. All detected labels, facial landmarks, and depth data are processed locally using Apple's Vision framework and CoreML. No cloud. No tracking. No subscriptions.