Test 001 · Transformation · 2026-10-13

One product photo → AI commercial

Input: one ordinary photo of a perfume bottle on a kitchen counter. Tools: Z-Image-Turbo (sets + relighting) and Wan 2.2 (motion), both free, Apache-2.0 licensed and running locally in ComfyUI. Output: a 9-second commercial cut from 4 AI shots, plus the shot that failed.

The commercial. Sets and motion are AI-generated from the one photo. Cuts, music placement and the title card are our edit.

The input

The single input photo: a perfume bottle on a white kitchen counter

"HALVEN" is a fictional brand. To avoid using anyone's real product, this input photo was itself made with an AI image model (Z-Image-Turbo), as a deliberately ordinary phone-style shot.

Verdict

7.5/10

Label text stays exact in every shot (original pixels)Convincing smoke, fire, sand and water motionFree, Apache-2.0, runs offline, no watermark Orbit shot: the label duplicates onto a second face (~3 s in)Needs a compositing step + ~7 min per shot on an RTX 5090Water: ripples and drops, but no real splash

From one ordinary photo, a fully Apache-licensed pipeline (Z-Image builds and relights the set, Wan 2.2 adds motion) produced dolly-ins, smoke, fire and sand shots that look commercial-grade, and the label never morphs because it's the original photo. It breaks when the camera circles the product: in the orbit shot Wan paints the label onto a second face of the glass. Use static and push-in shots; avoid orbits.

How it was made

1. Cut-out. The bottle is cut out of the one photo with a geometric mask (a script, no AI). The label pixels are kept exactly as photographed.

2. AI set. Z-Image generates an empty set (rock, water, dune, fire, marble) from a text prompt. The bottle is placed on it with a contact shadow.

3. AI relight. A masked Z-Image pass relights the glass and liquid to match the set. The label is protected and pasted back unchanged.

4. AI motion. Wan 2.2 animates that frame from a one-sentence camera brief: 81 frames, 720×1280.

Every shot, exactly as generated

Shot	Motion prompt (Wan 2.2, verbatim)	Seed	Time
Smoke on volcanic rock 81 frames · 720×1280 · 8 steps	Slow smooth cinematic dolly-in toward the perfume bottle standing on the wet black volcanic rock. Thin wisps of smoke curl slowly around the bottle, the warm amber backlight glows through the liquid, subtle light flicker on the wet rock. The perfume bottle stays perfectly rigid and unchanged: same shape, same cap, same label, no morphing. Photorealistic luxury fragrance commercial, cinematic. Set prompt: Luxury fragrance advertising photograph, empty set before the product is placed. A wet black volcanic rock with a flat top surface in the center foreground at eye level, the flat top occupies the middle of the frame. Thin wisps of white smoke curl slowly in the dark air, warm amber rim light from behind, tiny water droplets on the rock, deep black studio background, shallow depth of field, cinematic, photorealistic, high-end commercial. Nothing on the rock, no bottle, no text.	1102	6.6 min + set 9 s, relight 9 s
Water 81 frames · 720×1280 · 8 steps	Locked-off camera with a very slow push-in. Clear water droplets fall from above around the perfume bottle standing in the shallow black water, creating crisp crown splashes and expanding ripples around its base, slow motion look, warm amber backlight glow, reflections in the water. The perfume bottle stays perfectly rigid and unchanged: same shape, same cap, same label, no morphing. Photorealistic luxury fragrance commercial, cinematic. Set prompt: Empty dark still-life photography set. A glossy black floor covered by a thin layer of perfectly still black water, mirror reflections, the waterline horizon sits slightly below the middle of the frame, deep black background with a soft warm amber glow in the center behind, a few concentric ripples in the center foreground, cinematic, photorealistic. The set is completely empty: no objects, no bottle, no product, no text.	1202	6.6 min + set 6 s, relight 6 s
Desert sunset 81 frames · 720×1280 · 8 steps	Slow cinematic crane-down toward the perfume bottle standing on the golden sand dune at sunset. Fine sand blows gently across the dune in the wind, the low sun behind glows through the amber liquid, heat shimmer in the distance. The perfume bottle stays perfectly rigid and unchanged: same shape, same cap, same label, no morphing. Photorealistic luxury fragrance commercial, cinematic. Set prompt: Empty still-life photography set in the desert. The smooth crest of a golden sand dune in the center foreground at eye level with a small flat spot on top, fine wind ripples in the sand, low setting sun behind the crest creating a warm glow and long shadows toward the camera, endless soft dunes out of focus behind, cinematic, photorealistic. Completely empty: no objects, no bottle, no product, no text.	1301	6.6 min + set 6 s, relight 6 s
Embers hero 81 frames · 720×1280 · 8 steps	Very slow cinematic push-in toward the perfume bottle standing on the charred log. Glowing embers and bright sparks float upward around it in slow motion, flames flicker and lick behind the bottle, warm smoky atmosphere, the liquid glows like honey. The perfume bottle stays perfectly rigid and unchanged: same shape, same cap, same label, no morphing. Photorealistic luxury fragrance commercial, cinematic. Set prompt: Empty still-life photography set. A flat piece of charred black wood lies in the center foreground at eye level as a small stage, surrounded by glowing amber embers and small flames, sparks floating upward, warm smoky dark atmosphere, lots of dark negative space above, cinematic, photorealistic. Completely empty: nothing standing on the wood, no bottle, no product, no text.	1501	6.6 min + set 6 s, relight 6 s
Marble orbit (the flaw) 81 frames · 720×1280 · 8 steps	The camera slowly orbits around the perfume bottle on the black marble pedestal in a smooth half circle, revealing its side and thick glass edges, soft light streaks glide across the glass, reflections on the marble. The perfume bottle stays perfectly rigid and unchanged: same shape, same cap, same label, no morphing. Photorealistic luxury fragrance commercial, cinematic. Set prompt: Empty still-life photography set. A polished black marble pedestal block with white veins in the center at eye level, its flat top surface empty, minimal dark studio, a soft diagonal light streak across the background, glossy reflections on the marble, cinematic, photorealistic. Completely empty: nothing on the pedestal, no bottle, no product, no text.	1401	6.6 min + set 6 s, relight 6 s

Time is wall-clock from submit to file on one RTX 5090 (32 GB). Each motion shot was generated once (seed listed) and used as-is; the orbit shot appears only as the flaw. Set and relight passes used 2 seeds each, best one picked by eye.

Can you do this?

You need

✓ A GPU with lots of VRAM (we used 32 GB)

✓ ComfyUI + Wan 2.2 I2V + Z-Image-Turbo (all free)

✓ ~7 minutes per 5-second shot

No big GPU?

Cloud tools do photo-to-ad in a browser. Next we run this exact photo through Invideo AI, and the results go up here.

See the Invideo test file

Licenses

Wan 2.2 (I2V A14B), its text encoder, VAE and Lightning LoRA, plus Z-Image-Turbo, its text encoder and VAE, are all Apache-2.0. We checked the licenses and verified that our local files match the published ones. Music: ACE-Step 1.5 (MIT), generated locally. Fonts: SIL OFL.

All footage on this page is AI-generated, and the input photo is AI-generated too, showing a fictional brand. No sponsorship: we earn nothing from these tools.