Window Peeper AI Video Examples and Test Method

Window Peeper AI examples must pair a real test input with its complete generated output.

The reviewed examples below show permission-cleared input portraits and complete test videos. Each card identifies its own test version, measured output and observed limitations; selected results do not establish a universal success rate.

Watch the reviewed Window Peeper AI video examples

Our portrait-to-window test

One AI-synthetic adult portrait, one paid generation. Push-in, burger bite, stare and full curtain closure are visible; the exterior is very dark. This original model output is silent. One result does not establish a success rate.

Recipe status

Historical test; does not verify the current recipe

Tested

2026-10-23 · 14.042s · 720 × 1280 · 9:16 · none

Recipe and prompt

window-peeper-v1 / window-peeper-prompt-v2

Tested provider and model

apimart / seedance-2.5

Recorded run time (seconds)

417

Actual supplier cost for this run

USD 3.033

Recorded attempts in the same test group

1

Observed limitations

One paid generation. Reviewed frames show the camera approach, burger bite, stare, burger placed down and full curtain closure. Hair and clothing were broadly retained; the building exterior is very dark and less detailed than the references. Raw output is silent. A single selected result does not establish stable likeness or success rate. The complete input and original MP4 were archived and verified without another generation.

Historical supplier cost, not today’s selling price. Other attempts are not included in this run’s cost.

Published with explicit permission and editorial review. Permission records remain private.

What a Window Peeper AI video example must show

We require at least one reviewed single-person result from our own current recipe before opening generation. One portrait test demonstrates that result only; other appearances, profiles and difficult inputs may behave differently. External videos on the homepage are separately credited references and do not count as our own tests. If a test fails, it remains in the private test history; another successful result does not erase it.

Each reviewed example identifies its test date, recipe and prompt version, actual duration, pixel dimensions, audio, input conditions and visible limitations. The corresponding input and full MP4 must come from that same test. Quality notes distinguish a technically completed file from faithful burger, stare and curtain action. A successful API task alone is insufficient.

How to read a Window Peeper AI video example

Compare the input with the person after the camera approaches, then watch through the last frame. Look for the same hair and clothing, a readable bite, a deliberate stare, the burger being released, and the window being covered. Read the limitation beside the player before deciding whether the scene suits your own idea.

The v2 prompt record documents one October 23 test: a 14.04-second, 720 × 1280 silent file whose exterior was visibly dark. That observation belongs to one synthetic adult portrait. It is not a customer testimonial, a promise about your face or evidence that paid delivery is open.

What we inspect in the whole clip

The camera must begin outside at night, approach a lit window, show the adult eating a burger, let them notice and stare into the camera, then end with the curtains fully closed. We inspect identity through camera movement, coherent hands and food, shot continuity, audio, and the actual saved download. No percentage success or likeness score is claimed from a selected test.

Use each card’s recorded specifications when comparing results. Earlier tests retain their original recipe, prompt and measured output; changing the current product does not turn them into new results.

A short result cannot be relabeled as a complete recreation. When comparing Window Peeper AI examples, distinguish a complete silent base video from an edited version with music. An added soundtrack does not verify the original audio rights or improve the measured identity consistency.

How permission and withdrawal work

Account uploads and outputs are private by default; only separately approved examples are public. Public selection needs explicit permission covering the displayed portrait and result, plus review. Permission references and private asset identifiers are not published. Withdrawing a selected example removes it from public queries and its media endpoints; downloaded copies cannot be recalled. Public sharing of account videos is disabled.

Comparing models or fixing a failure

A comparison would use the same authorized input and target scene, and record actual parameters, start and completion times, attempt grouping, costs and observations for the camera, face, hands and curtains. Different conditions must be disclosed. Unknown costs stay unknown; provider price estimates do not become historical test spend. A model ranking needs comparable results, and a claimed repair needs a published retest; a suggested fix alone is not validation.

Prepare a portrait with the guide · Read the candidate prompt · Check pricing and availability