Sogni 360 & Montage

Two cinematic modes in one app. Montage stitches your photos and clips into epic travel and event recaps. 360 turns a single still into a smooth orbital video. Together they make any moment feel directed.

#Sogni Montage

Sogni Montage is the travel-video and highlight-reel mode of Sogni 360. You bring a stack of stills from a trip, an event, or a campaign and arrange them in order; Sogni generates short AI Transitions between every pair of frames so the stills play as continuous motion. You set the clip duration on each scene — that's where pacing comes from — and re-roll any scene with prompt feedback until the motion lands the way you want. The Supernet runs the renders in parallel batches so a full montage comes back in one pass instead of one cut at a time.

It's the fastest way to take a camera-roll dump from a weekend (or a two-week vacation) and turn it into something people actually want to watch.

#Featured Montages

Two weeks across the South Island — fjords, glaciers, road-trip diners — distilled into a 45-second cinematic recap via Sogni 360's Montage transitions feature. Made on a phone from 23 static trip photos plus 2-second AI Transitions generated inline in the workflow.

Open this sample project →

#Sogni 360

Sogni 360 turns a single still image into a seamless 360-degree orbital video. Upload a photo, pick the angles you want, and the Sogni Supernet re-renders the same subject from up to 96 camera positions, then stitches AI-generated transitions between them. No 3D scan, no rig, no GPU.

#Launch Sogni 360 →

Reviewing a snowboarder rendered from multiple camera angles in Sogni 360

#What it does

96 camera angles from a single source image — 8 azimuths (front, front-right, right, back-right, back, back-left, left, front-left) × 4 elevations (low-angle -30°, eye-level 0°, elevated 30°, high-angle 60°) × 3 distances (close-up, medium, wide). Up to 64 waypoints per project.
Qwen Image Edit 2511 + Multiple Angles LoRA for the angle re-renders — Lightning tier for fast scouting, full 2511 for higher fidelity. Subject identity, lighting, and clothing stay consistent across the orbit.
Two transition model families, switchable per project:
- WAN 2.2 — proven quality, four steps (Fast and Balanced run the LightX2V distilled variant; High Quality and Pro run the full WAN 2.2 14B). 16fps base, post-interpolated to 32fps. 1–8s per clip.
- LTX-2.3 — next-gen model with synchronized audio generation (foley + ambient, woven into the prompt automatically — not music). Native 24fps, up to 20s per clip, longer durations than WAN. Minimum 640p.
Resolutions 480p, 580p, 640p, 720p, 1080p (LTX-2.3 starts at 640p).
AI-expanded transition prompts. Pick a motion style; a Sogni Intelligence VLM analyzes the two end frames and rewrites the preset into a scene-aware, model-tuned prompt (the LTX-2.3 path bakes in the required audio cues automatically).
Sound-to-video on the final stitched video. Pick a track from the preset music library to lay over the assembled orbit.
Batch generation of every angle in parallel, with per-angle regeneration when one comes back off.
Full-screen immersive viewer with click zones, swipe, keyboard shortcuts, and auto-play.

Sogni 360 is built on the Sogni Client SDK. The same 3D camera-angle remap powers the angle remix feature inside Sogni Photobooth for Web.

#Preset templates

Two layers of presets so you don't dial settings from scratch.

Camera-angle templates seed a project with a curated set of waypoints:

Simple Zoom Out (2 angles) — original + zoomed-out wide.
Zoom Out 360 (9) — full eye-level rotation in 45° steps, returning to the start frame for a clean loop.
Zoom Out 360 (4) — same idea in 90° steps for a faster, choppier orbit.
Zoom Montage (4 angles) — original plus three dynamic angles mixing high, low, and wide for a more cinematic cut.
Portrait Trio (4 angles) — original, three-quarter left and right, and an elevated front shot. Designed for character sheets and portrait turnarounds.
360 Spin Close (9) — close-up eye-level orbit in 45° steps.
360 Spin High+Wide (9) — high-angle wide orbit for landscapes, architecture, and product hero shots.
360 Spin Random (9) — varied elevations and distances for documentary or storytelling cuts.

AI Transition prompt presets set the motion style the video model targets between waypoints:

Camera Orbit (default) — smooth orbital motion between angles of the same scene. Best for orbit and turnaround templates.
Cinematic Morph — dynamic on-lens transition with brief natural occlusion. Best when the two end frames are different photos (Montage-style cuts).
Parallax Zoom — epic depth-layered dolly-zoom reveal, foreground sliding past, midground drifting, background steady.
Parallax Zoom 2 — cinematic pull-back variant of Parallax Zoom with softer easing.

Either preset can be overridden inline — write your own prompt and the VLM still expands it scene-aware against the two end frames.

#360 workflows

Action sport hero shot. Upload a clean freeze-frame — a BMX rider mid-trick, a surfer in the barrel — pick Zoom Out 360 (9), and you get a loopable orbit you can export for Instagram or TikTok. Switch to LTX-2.3 for the export if you want crowd noise, wind, or wave foley baked in.

Vehicle reveal. Drop in a single car photo, run a 360 Spin High+Wide for the establishing pass, then a Camera Orbit at eye-level for the hero loop. WAN 2.2 Pro for client-deliverable cuts; LTX-2.3 when the deliverable needs engine and wind ambience.

Character turnaround. Use Portrait Trio on a Photobooth portrait to build a consistent character sheet without rerolling identity. Cinematic Morph between the front and three-quarter frames gives the sheet a film-still feel.

Architecture and product spins. Generate the angle batch from a hero render, regenerate any individual angle that drifts, then export the orbit as a video loop for product pages, listings, or 3D-style previews. Add a music preset for social-cut versions, or LTX-2.3 if you want the space's room tone to play under the spin.

#Tips

One clear subject. The Multiple Angles LoRA works best when the subject is well separated from the background — busy scenes can confuse the re-projection. Lightning tier is fine for scouting; switch to the full Qwen Image Edit 2511 for final angle renders that need to hold up close.
Pick the model family for the deliverable, not just the speed. WAN 2.2 wins on raw motion quality and is the default; LTX-2.3 wins when you need synchronized audio in the output, longer clips (up to 20s), or native 24fps without post-interpolation.
Let the AI expand the prompt. The preset prompt is fine, but the VLM-expanded version reads both end frames and writes targeted camera, lighting, and (on LTX-2.3) audio cues. The expanded prompt is editable before render — tune it once and reuse across the orbit.
Generate the still batch first, transitions second. Cheap to explore angles, expensive to render video — lock the angles you like before spending steps on transitions.
Regenerate, don't restart. If one angle or transition comes back wrong, regenerate just that waypoint instead of rerunning the whole orbit.
Match the transition preset to the source. Use Camera Orbit when adjacent waypoints are AI-generated angles of the same scene; switch to Cinematic Morph for Montage-style cuts between different photos.
Use Fast for scouting, Pro for the final loop. Same workflow pattern as Photobooth — explore on Fast (LightX2V on WAN 2.2, or LTX-2.3 Fast for audio previews) and finalize on Pro.