Kling virtual try-on API: what it takes and returns

Kling's virtual try-on is an image API: you POST one garment image and one person image to /solutions/virtual_try_on, and it returns a try-on image asynchronously, not a video. Kling released Virtual Try-On 3.0 on 09/15/2026; it had supported the earlier Kolors try-on API (kolors-virtual-try-on) since 09/19/2024, and says V1 and V1.5 parameters were adapted to the 3.0 pipeline.

Every Kling fact below comes from Kling's own API pages listed under Sources, read on 2026-09-28. The video steps use Sume's Format catalog and Video generation docs.

What does Kling's try-on request contain?

The body has one required array and two optional objects. contents must hold exactly one product_image and one person_image, each with a publicly accessible url. Three preservation switches, all on by default, decide what is kept from the person photo:

What images does Kling's virtual try-on accept?

Both images share the same file rules:

• JPG, JPEG, PNG, or WebP, at most 10 MB.
• The longest side at most 2048 px; the shortest side greater than 300 px.
• Garment: an on-person photo, a mannequin shot, a flat lay, or a product photo on a white or clean background.
• Person: one person, facing front or three-quarter, with the try-on area fully visible. With keep_pose on, Kling recommends a natural standing pose with the hands unobstructed.

How do I get the result, and what does it cost?

The create call returns a task_id and a status: submitted, processing, succeed, or failed. Poll GET /solutions?task_ids=… (up to 20 IDs, comma-separated) or use your external_task_ids, or wait for the callback. On success, the image is the url of the outputs entry whose type is image. Kling says generated images are cleared after 30 days, so save them promptly.

Kling's E-Commerce pricing page lists Virtual Try-On per image at 1K: 32 Units ($0.112) per image.

What changed between Kolors V1, V1.5, and 3.0?

Kling's changelog gives three dates. On 09/19/2024 it began supporting the “AI Virtual Try-on” API, kolors-virtual-try-on. On 12/30/2024 came the V1.5 model, an upgrade of V1.0 that supports single-garment try-on (upper, lower, and dress) and “upper + lower” combinations. On 09/15/2026 Kling released Virtual Try-On 3.0, which it says supports flat-lay, mannequin, and on-model garment images and the face, pose, and background controls above.

Can Kling's virtual try-on make a video?

Not by itself: its output type is an image. Kling lists a separate video service, AI Apparel Replicator (clothing_dupe), which replaces the garment in a reference video you supply and returns a 5–180 second showcase video.

To turn a try-on still into motion on Sume, host the image at your own public HTTPS URL and send it to POST /v1/videos as a frame_images entry with frame_type first_frame. To go from photos straight to video, Sume's catalog Formats sume-virtual-try-on and sume-virtual-fitting make a finished try-on video from a person photo and a garment photo; the call is in Virtual try-on video API. Which Kling models Sume runs is covered in Sume vs Kling.