Scheduled Maintenance

We're performing scheduled maintenance. Some features may be temporarily unavailable.

Choose Theme

JollyAI Default

Clean and professional dark theme

It Comes at Night New

Ultra dark JollyAI theme

Stranger Things

80s horror-sci-fi aesthetic

Batman

Dark knight, dark theme

Barbie

Bright and colorful Barbie theme

Ocean Blue

Twitter-inspired blue theme

Midnight Purple

Catppuccin-inspired purple theme

Forest Green

Matrix-style green theme

Crimson Red

Warm red-based dark theme

Amber Glow

Warm amber/orange theme

Dracula

Popular gothic-inspired theme

Monokai

Classic Sublime Text theme

Nord

Arctic-inspired cool theme

Gruvbox

Community favorite warm theme

Solarized Dark

Carefully calibrated for readability

One Dark Pro

Popular VS Code theme

Wan 2.2 + VACE Seamless Loop Video Workflow for ComfyUI

Turn one image into a perfectly seamless looping video — 1600p, 32fps, any length you set in seconds. The full technique behind the VACE loop-closing pass, every stage of the pipeline, and the exact production workflow we deliver to clients (70 nodes), available below for $3.99.

Get the Workflow — $3.99 Try Wan 2.2 Free on JollyAI

From one image to a seamless 1600p loop

Drop in a single still and the workflow returns a perfectly seamless looping video at up to 1600p, 32fps, in any length you choose. Here's a real input image and the exact looping output it produced — the clip below is the unedited result, playing on a loop so you can see there's no pop or jump at the seam:

Input still image for the Wan 2.2 + VACE seamless loop ComfyUI workflow
Input — a single 1024×1024 still image
Output — seamless looping video · 1600p-ready · 32fps · variable length
1600p resolution 32 fps (RIFE) Variable length — set seconds Mask — animate only what you want Truly seamless — no pop Single 16 GB GPU

Why "seamless loop" is the hardest AI video problem

Looping videos are everywhere — live wallpapers, stream overlays, product ads, lo-fi backgrounds, NFT animations, digital signage. They also sell: a clip that loops invisibly reads as "professional" in a way a normal AI clip never does.

The problem: image-to-video models like Wan 2.2 only condition on the first frame. The model knows where the video starts but has no idea (and no incentive) to land back there. The last frame always drifts — lighting shifts, the subject moves, the camera floats — so when the player jumps back to frame one you get a visible "pop".

Crossfading the ends in an editor doesn't fix it; it just smears the pop over 10 frames. The real fix is to make the model generate a transition that ends exactly on the first frame. That's what VACE (Video All-in-one Creation and Editing) makes possible — it accepts reference frames as hard constraints, so you can demand: "start from the clip's last frames, end on its first frames."

How the VACE loop-closing technique works

  1. Pass 1 — base motion (Wan 2.2 I2V): a normal image-to-video generation using Wan 2.2's dual-expert setup — the high-noise expert blocks out motion, the low-noise expert refines detail. This produces your main clip.
  2. Pass 2 — the bridge (Wan 2.1 VACE): the workflow extracts the tail frames of the clip and its opening frames, reverses and batches them as VACE reference context, then generates a short bridge segment that starts where the clip ended and lands precisely on frame one.
  3. Composite: base clip + bridge are concatenated with the overlap trimmed and masked, so there is no duplicated frame and no seam.
  4. Polish: RIFE interpolation doubles the frame rate to 32 fps and the result is muxed by VHS Video Combine. The loop point becomes mathematically invisible — the last frame's successor is the first frame.

The trick that makes the bridge look natural rather than "morphy" is feeding VACE several context frames from each end (not just two stills) — it learns the motion direction in and out of the seam, so momentum carries through the loop point.

Inside the production workflow (70 nodes)

This is the exact pipeline of the workflow you can download below — the same file we deliver on Fiverr:

Stage Key nodes What it does
Auto-captioningFlorence2Run + StringConcatenateDescribes your input image automatically and builds the motion prompt for you — no prompt writing needed.
Base generationWanImageToVideo, 2× KSamplerAdvancedDual-expert Wan 2.2 I2V pass (high-noise → low-noise) for the main clip.
Speed-uplightx2v step-distill LoRA, PatchSageAttentionStep-distilled sampling + SageAttention — cuts generation time roughly in half on consumer GPUs.
Loop bridgeWanVaceToVideo, ReverseImageBatch, ImageFromBatchBuilds the VACE reference stack from both ends of the clip and generates the seam-closing segment.
Mask / area controlImageCompositeMasked, ImageToMaskComposite the animated subject over a frozen copy of the image through a mask — only the masked region moves, everything outside it stays perfectly still (ideal for animating water, sky, hair or eyes while the rest holds).
CompositingBatchImages, exact frame accountingConcatenates base clip + loop bridge with no duplicated frame and no seam.
Length controlDuration (seconds) inputType how many seconds you want — the frame count scales automatically for a 5s, 10s or longer loop.
Upscale, polish & exportSeedVR2VideoUpscaler, RIFEInterpolation, VHS_VideoCombineUpscales to 1600p, interpolates to 32 fps and exports the final seamless loop.
VRAM hygieneeasy cleanGpuUsed / clearCacheAllFrees VRAM between passes so the whole pipeline runs on a single 16 GB card.

Models you'll need (all free downloads)

Model Role
Wan 2.2 I2V high-noise + low-noise (fp8 / smoothMix merges)Base image-to-video generation
Wan2.1 14B VACE (Q8 GGUF)Loop-closing bridge pass
lightx2v cfg-step-distill LoRA (rank 32)4–8 step fast sampling
umt5_xxl fp8 text encoder + Wan 2.1 VAE + clip_vision_hConditioning & decode
Florence2 (auto-downloaded by the workflow)Automatic image captioning

The download includes a notes panel inside the workflow with the exact filenames and where each model goes (diffusion_models, loras, text_encoders, vae, clip_vision). Custom node packs used: ComfyUI-VideoHelperSuite, KJNodes, rgthree, ComfyUI-Easy-Use, Florence2, RIFE, LayerStyle — all installable via ComfyUI Manager.

Get the complete workflow — $3.99

  • Production-tested 70-node ComfyUI workflow JSON (drag & drop to import)
  • 1600p · 32fps output with SeedVR2 upscale + RIFE interpolation
  • Variable length — just set the loop duration in seconds
  • Mask control — animate only the area you paint, keep the rest perfectly still
  • Florence2 auto-captioning — works even if you can't write prompts
  • Runs on a single 16 GB GPU (RTX 4080 / 4070 Ti S class)
  • Instant download + backup link sent to your email

Secure payment via PayPal / Razorpay. The download link works 20 times. Questions? Email us.

FAQ

What exactly do I get for $3.99?

The complete 70-node workflow JSON — the same file we deliver to Fiverr clients — with embedded setup notes. Import it into ComfyUI via drag & drop.

What resolution, frame rate and length does it output?

Up to 1600p at 32 fps (SeedVR2 upscale + RIFE interpolation), and the length is yours to choose — set the loop duration in seconds and the workflow scales the frame count automatically, so you can make 5-second, 10-second or longer seamless loops from the same graph.

Can I animate only part of the image?

Yes — there's a mask stage built in. It composites the animated result over a frozen copy of your image, so only the masked area moves and everything outside it stays perfectly still. Load a black-and-white mask to animate just the water, sky, hair, eyes or a flame while the rest of the frame holds steady — great for cinemagraphs and ultra-clean loops. Prefer full motion? A single Use Mask toggle turns the mask off so the whole frame animates.

How much VRAM do I need?

16 GB is comfortable (VACE runs as Q8 GGUF, Wan 2.2 experts as fp8). 12 GB works at lower resolution/frame counts.

Why doesn't my normal Wan 2.2 video loop?

I2V only conditions on the first frame, so the final frame always drifts away from it. The VACE bridge pass is what forces the video to land back on frame one.

Do I need to write prompts?

No — Florence2 captions your image automatically and the workflow assembles the motion prompt. You can still override it.

Can I sell videos I make with it?

Yes, your generated videos are yours, including commercial use. Only reselling/redistributing the workflow file itself is not allowed.

I'd rather not set up ComfyUI — can you generate loops for me?

Yes — try free Wan 2.2 image-to-video on JollyAI (5 free generations daily, no signup), or order a custom loop from us directly.

Related guides & tools

Text-to-Video Seamless Loop Workflow · Wan 2.2 Complete Guide · Wan 2.2 Cinematic Prompts · Make AI Video from Image (Free) · WAN Image to Video · Free Image to Video Tool