This corrected draft now compares Grok Imagine Video 1.5 against WAN I2V on three portrait image-to-video tests that were manually rechecked before update.
Each sample below is shown in the same order: source → Grok → WAN. If a source and outputs did not stay relevant enough to the same scene, the source was regenerated and the outputs were rerun. That happened on the chef test, which now uses a cleaner source image and fresh outputs for both models.
- Test setup
- Overall result
- Test 1: watch product shot
- Test 2: chef behind rainy glass
- Test 3: astronaut atrium scene
- Final verdict
Test setup
- 3 portrait source images
- Same motion prompt for Grok Imagine Video 1.5 and WAN I2V on each test
- 5-second target clips
- Portrait framing throughout
- 480p reruns for both models on the final kept set
- Manual recheck of source relevance against both outputs before publishing this draft update
Overall result
| Model | Verified usable clips in this draft | Best at here | Main weakness |
|---|---|---|---|
| Grok Imagine Video 1.5 | 3/3 | Keeping scene identity intact, especially in the chef shot | Can look softer or slightly less grounded in hard-surface and cinematic environment shots |
| WAN I2V | 3/3 | Stronger realism in the watch and astronaut tests | Sometimes smooths away specific scene identity and drifts a bit more from the source |
WAN I2V wins this corrected draft by a 2-to-1 margin.
Test 1: watch product shot
Motion prompt: Slow macro orbit around the watch while cool mist drifts and tiny reflections slide across the metal surface. Premium commercial motion, steady camera, realistic detail.
| Source | Grok | WAN |
|---|---|---|
 |
|
|
The earlier watch row was rejected after visual QA because the embedded sample was clearly wrong. This rebuilt version now keeps both outputs on the intended watch product-shot scene. WAN still looks slightly cleaner in the reflections and metal finish, while Grok is now at least showing the correct subject.
Test 2: chef behind rainy glass
Motion prompt: Slow push in toward the chef as steam curls upward and neon signs glow through rainy window reflections. Natural blinking, subtle head movement, cinematic realism.
| Source | Grok | WAN |
|---|---|---|
 |
|
|
The earlier chef sample was not clean enough, so this row was rebuilt from a new source image. On the kept version, both outputs are relevant to the same chef-through-rainy-glass scene. Grok wins because it stays closer to the source composition and details, while WAN drifts more in pose and utensil action.
Test 3: astronaut atrium scene
Motion prompt: The camera starts behind the astronaut and arcs to the front as dust floats in skylight beams and a few vines sway gently. Cinematic realism, steady movement.
| Source | Grok | WAN |
|---|---|---|
 |
|
|
Both outputs stay relevant to the same astronaut-in-atrium scene. WAN wins because the suit, lighting, and environment feel more grounded, while Grok shifts the composition more noticeably.
Final verdict
After the relevance recheck, all three kept rows now pass the basic source-to-output test. WAN I2V wins the watch and astronaut rows. Grok Imagine Video 1.5 wins the rebuilt chef row.
That leaves WAN I2V slightly ahead overall on this verified three-test set, while Grok still looks stronger when exact scene identity matters more than cinematic polish.
