Virtual try-on / MERCHANT WORKFLOW
Should You Use FASHN Performance Mode Before Generating Final Try-On Images?
For FASHN Try-On v1.6, the vendor recommends performance mode when exploring model-and-garment combinations and quality mode for final image generation. Balanced is an intermediate option. This is a documented speed-versus-output-quality setting, not a guarantee that a performance candidate predicts the final result or that the overall campaign will cost less.
Choose the mode according to the unresolved question
FASHN's Try-On v1.6 documentation defines mode values performance, balanced, and quality. Performance prioritizes speed with reduced image quality. Balanced offers an intermediate trade-off. Quality is slower and intended for higher-quality output.
The vendor recommends performance for identifying useful model-and-garment combinations before requesting final results in quality mode.
The key distinction is what the merchant is trying to learn. During exploration, the question may be whether the model pose leaves the garment visible, whether the input photographs are suitable, or whether the intended garment category is recognized. At final review, the question becomes whether the actual delivered file is accurate and usable for its destination.
Do not infer that a performance image is an exact low-resolution preview of a later quality image. The later request is a separate generation whose result must be reviewed independently.
The current documentation also distinguishes Try-On v1.6 from Try-On Max. These exact mode values should not be generalized to every FASHN endpoint or to CartVisual.
Keep input and variation choices consistent
For a controlled editorial comparison, record the same model_image, garment_image, garment category, garment photo type, and intended composition across the proposed requests.
FASHN documents garment_photo_type values auto, model, and flat-lay. It also documents garment categories such as tops, bottoms, and one-pieces. These input interpretations can affect the task independently of the generation mode.
The num_samples parameter can request multiple variations. FASHN suggests two to four performance-mode samples for exploring seed sensitivity. More samples are additional generated candidates, not free duplicates or a measured probability of success.
Record the current account quote and actual billing separately for each submitted operation. The vendor's speed recommendation does not establish a universal monetary advantage, especially if the merchant later requests several additional generations.
The existing resolution-credit guide
Hypothetical cardigan screening and final request
Hypothetical exercise. No FASHN job was submitted, timed, billed, or compared. The garment is CARDIGAN-FOREST-07, with source cardigan-forest-front-v3.png. The target model image is model-neutral-standing-v2.jpg.
The proposed screening request TRYON-SCREEN-01 uses mode: performance, category: tops, garment_photo_type: flat-lay, and num_samples: 2. The merchant intends to inspect whether the front buttons and sleeve shape remain identifiable.
Only if a suitable combination is identified does the merchant consider TRYON-FINAL-02, using mode: quality and one candidate for detailed review. The final request is not an automatic enhancement of the screening file.
| Request | Mode | Samples | Decision supported |
|---|---|---|---|
| TRYON-SCREEN-01 | performance | 2 | Whether the input combination merits further work |
| TRYON-FINAL-02 | quality | 1 | Actual final-file review after separate generation |
No mode has been declared a quality winner or cost winner.
Stop exploration when the source problem is not a mode problem
If screening candidates consistently obscure the cardigan's defining button row, reconsider the model pose, garment source, or intended framing before paying for a final request. Quality mode cannot establish construction hidden by inadequate evidence.
If one performance candidate looks promising, that supports considering the next step; it does not approve the future quality output.
If the final candidate changes the cardigan's collar or buttons, hold the image regardless of its technical detail. Compare the garment against the approved source and consider a source-preserving alternative.
If the apparent problem is remnants of the model's original clothing, use the symptom-specific diagnosis in the segmentation-free guide rather than assuming the generation mode is responsible.
Review actual job status, outputs, and billed usage before recording a production decision. CartVisual does not expose FASHN's mode parameter or execute its API requests.
Use the example with your own product records
The filled records are hypothetical editorial examples. The entry-template row is for your own files, settings and observations; it does not connect to another application.
| Request | SKU | Model image | Garment image | Mode | Category | Photo type | Samples | Actual quote | Actual output | Next action |
|---|---|---|---|---|---|---|---|---|---|---|
| TRYON-SCREEN-01 | CARDIGAN-FOREST-07 | model-neutral-standing-v2.jpg | cardigan-forest-front-v3.png | performance | tops | flat-lay | 2 | Not quoted | Not generated | Review input compatibility and garment visibility |
| TRYON-FINAL-02 | CARDIGAN-FOREST-07 | model-neutral-standing-v2.jpg | cardigan-forest-front-v3.png | quality | tops | flat-lay | 1 | Not quoted | Not generated | Submit only after screening decision and review final independently |
What the sources establish
- FASHN — Try-On v1.6 Parameters: Documents performance, balanced, and quality modes, exploration recommendation, garment photo type, categories, sample count, and seed behavior.
Sources and further reading
Recorded source-review dates are listed individually below. An undated source does not inherit the article’s publication date. Software and platform rules can change; check the current original before acting.
- FASHN — Try-On v1.6 Parameters · Review date not separately recorded
Frequently asked questions
Does FASHN performance mode guarantee a cheaper approved final image?
No. It is documented as a faster exploration setting with reduced image quality. Total cost depends on actual requests, current billing, and the number of usable results.
Can I approve a quality-mode result based on an earlier performance-mode preview?
No. A quality-mode request produces its own result and requires independent product and destination review.
YOUR NEXT CAMPAIGN
Create assets from a clear source
Bring your reviewed source photo or brand brief to CartVisual. Check the current tools, account access and credits before generating, then compare every result with your source.
Create a CartVisual account Open the workspace Check access and credits