{"id":1225,"date":"2026-02-27T18:51:46","date_gmt":"2026-02-27T18:51:46","guid":{"rendered":"https:\/\/wiro.ai\/blog\/?p=1225"},"modified":"2026-09-27T16:51:43","modified_gmt":"2026-09-27T16:51:43","slug":"avatarmotion-multi-6-two-photo-animations-tested","status":"publish","type":"post","link":"https:\/\/wiro.ai\/blog\/avatarmotion-multi-6-two-photo-animations-tested\/","title":{"rendered":"AvatarMotion Multi: 6 Two-Photo Animations Tested"},"content":{"rendered":"

AvatarMotion Multi<\/strong> turns two supplied photos into a short preset animation. This test keeps the same two portrait inputs and seed across six runs, then changes only effectType<\/code>. That makes the clips useful for checking a practical question: how much of the result comes from the chosen interaction rather than from a new prompt or a different pair of photos.<\/p>\n

AvatarMotion Multi on Wiro<\/a><\/p>\n

What this AvatarMotion Multi test checks<\/h2>\n

This is not a general image-to-video benchmark. The model’s documented interface asks for uploaded images rather than a text prompt, and the selected preset supplies the motion idea. For this post, the setup uses two people images in every run, seed=42<\/code>, and one of the six Polaroid interaction presets. The aim is to see whether an action stays legible while the two inputs remain recognizable.<\/p>\n

The official model documentation lists a broader set of two-image options too: relationship scenes, pet scenes, family-photo compositions, a car option, and a one-person-plus-pet ride preset. It also specifies when a preset needs two or three inputs. The six Polaroid presets below all call for two people images, so they are a clean like-for-like set. For context on the Wan 2.2 work listed in the model documentation, see the official Wan 2.2 repository<\/a>.<\/p>\n

Inputs and fixed parameters<\/h2>\n\n\n\n\n\n\n\n\n
Setting<\/th>\nValue<\/th>\nWhy it matters<\/th>\n<\/tr>\n<\/thead>\n
Inputs<\/td>\nTwo portrait photos<\/td>\nEvery Polaroid preset in this test documents two people images.<\/td>\n<\/tr>\n
Seed<\/td>\n42<\/td>\nHeld constant so the preset is the intended variable.<\/td>\n<\/tr>\n
Prompt<\/td>\nNone<\/td>\nThe effect type selects the action; no text prompt was added.<\/td>\n<\/tr>\n
Output choice<\/td>\nSix existing blog-hosted videos<\/td>\nEach clip is the saved result for one preset.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\n\n\n\n\n
Input photo A<\/th>\nInput photo B<\/th>\n<\/tr>\n<\/thead>\n
\n
\"AvatarMotion
Input A: first portrait used in all six runs.<\/figcaption><\/figure>\n<\/td>\n
\n
\"AvatarMotion
Input B: second portrait used in all six runs.<\/figcaption><\/figure>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Six AvatarMotion Multi outputs<\/h2>\n

These clips show preset-driven interactions, not six unrelated generations. The useful comparison is the clarity of the requested gesture and whether faces, arms, and the paired composition stay plausible while it happens. Hands and close contact are the hard part of this category, so the handshake, high-five, and victory-sign clips deserve closer inspection than a simple facial expression.<\/p>\n

1. Polaroid smiling<\/h3>\n

effectType=polaroid_smile<\/code>, seed=42<\/code>. This is the lowest-complexity test in the set: a paired portrait with a small facial-action cue. Use it when the desired result is a light social animation rather than a scene change. It is also the best first run for checking whether the source faces and framing suit the model before moving to hand-heavy presets.<\/p>\n