Keep the evidence beside the output
Open the real garment references, selected model anchor and generated output together. Note whether the reference is front, back or detail. A hidden area has no direct visual evidence.
Our practice red dress is photographed from the back. That allows comparison of visible features, but it leaves the front uncertain. A nice front result cannot settle that question by itself. Add a real front reference or keep the detail marked unknown.
Use the seven fitting points
Record a status and a reason for each point. Length and fit may need actual measurements or confirmation from the garment owner. The face is checked against the approved identity reference, while the clothing is checked against the real product.
The checklist is a decision aid. A total score can hide the issue that matters for a specific image, so preserve the individual observations.
| Point | What to compare |
|---|---|
| Length | Hem and intended garment length; use measurements when needed |
| Sleeves | Length, shape and edge details |
| Neckline & collar | Visible shape and construction |
| Closures & details | Buttons, zips, trims and decorations |
| Fit & waist | Waist placement, silhouette and visible fit |
| Color & print | Color impression, pattern and placement |
| Face | Identity compared with the approved model anchor |
Record uncertainty instead of filling it in
Use pass when the evidence supports the feature, fail for an observed mismatch, unknown when evidence is missing and n/a when the point does not apply. Explain the observation with a short note.
An unknown closure stays unknown even if the rest of the image looks convincing. A fabric label is a better source for composition than the visual appearance of a generated dress.
Compare this output with my real garment photographs and model anchor. Review Length; Sleeves; Neckline and collar; Closures and details; Fit and waist; Color and print; Face. For each, record pass, fail, unknown or n/a, the evidence and the next action. Keep owner-confirmed measurements separate from visual estimates. Do not assume hidden details pass.
Add five checks when the result moves
A video can start well and drift later. Review the full clip at normal and slower speed. Listen to its audio. Add a time or interval to each observed issue so the next action is clear.
Use sampled frames to locate a change, then return to playback. A contact sheet alone cannot establish that the movement looks natural across every frame.
| Motion check | Look for |
|---|---|
| Fabric & opacity | Changing texture, transparency or hem behavior |
| Movement & contact | Foot slides, body drift and unnatural weight transfer |
| Camera & frame | Unwanted movement, missing shoes or cropped clothing |
| Background | New objects, distorted architecture or disappearing elements |
| Light & color | Exposure changes, darkening garment areas and color drift |
Make a decision for the intended use
Keep when the full reviewed result suits its purpose. Correct when an observable issue has a focused next step and a reasonable cost. Discard when the output cannot support the intended use or the next attempt has no clear direction.
A clip may be useful as an illustrative creative study while remaining unsuitable for a faithful product claim. Keep that limitation with the file. Do not let an attractive overall impression erase a failed detail.
Save the review with the final file
Keep originals, attempts, prompt, settings, reference links and review in the project folder. Label exports with their actual status. Retain unknowns and outstanding actions so the next person sees what remains to be checked.
After a correction, repeat the full review. A changed hem can come with a changed face, sleeve or light. Treat every new generation as a new output that needs evidence.
Questions before you start
What if I cannot see a detail?
Mark it unknown and record the missing evidence. Take another real reference photo when the detail matters.
Is seven out of seven a certification?
No. Keep the individual observations and source limits. The checklist is a practical review process, not independent product certification.
Can Claude perform the review for me?
It can help organize observations and inspect what is available to it. Ask it to state its limits, and confirm full video playback and the relevant real product details yourself.
Examples and sources
Adapted from the shared fashion-fitting-check method and the Motion Check in Add Video to Your Campaign.
Written by Fashion AI Skool using its course materials and practice records. Updated October 20, 2026. Generated examples demonstrate a process; they do not certify unseen product details.
