CartVisualCreate account

How should I configure FASHN Image to Video for a product-fashion clip without overspending or losing the job result?

Configure FASHN Image to Video in two passes: first decide whether the clip needs automatic motion or a specific end frame; then select the least expensive duration and resolution that answer the review question. Use one valid starting image, keep any motion instruction short, and retain the returned prediction ID with the exact configuration. Treat the generated clip as creative output for review, not standalone proof of physical product accuracy.

Published · By CartVisual editorial team

Define the transition

Use the starting image alone when the goal is to animate a still without prescribing a final pose. Leave prompt empty for automatic motion; if direction is necessary, use one short physical action rather than a storyboard. FASHN says detailed prompting is difficult to control precisely.

Add end_image only when the final composition is part of the deliverable. It makes the request a start-to-end transition, so compare the two source frames before submission: subject identity, garment appearance, crop and camera angle should support the intended change. If there is no required final frame, omitting the field keeps the decision simpler.

Before choosing output settings, check both input images against the documented limits. An image must be at least 301 × 301 pixels and no larger than 30 MiB. Resolution selection does not replace input validation, and FASHN does not upscale an undersized source.

Buy only the detail you need

The controlled escalation is therefore directional: hold resolution while testing timing, or hold duration while testing detail. For example, moving from 5 seconds at 720p to 10 seconds at 720p changes the listed cost from 3 to 6 credits; moving instead to 5 seconds at 1080p also changes it from 3 to 6. Those changes answer different questions.

ConfigurationListed creditsDecision use
5s / 480p1Motion proof when fine detail is unimportant
5s / 720p3Balanced first review candidate
5s / 1080p6Short candidate requiring more detail
10s / 480p2More time with minimal resolution cost
10s / 720p6More readable action at moderate detail
10s / 1080p12Longest, highest-detail listed option

Preserve the job state

The documented request uses POST /v1/run, with model_name set to image-to-video and endpoint parameters inside inputs. A successful submission returns a prediction ID rather than the finished video. Store that ID beside the source asset references, prompt state, duration, resolution, and end-frame choice.

Polling is a separate state transition: submitted is not completed. Continue through the documented status workflow until the response reports completion or an error. On completion, the response includes an output array with the generated MP4 URL. Retain the returned result according to your own asset policy, then review the MP4 against the approved source and intended motion before publishing.

The practical control is a small job record: configuration at submission, prediction ID immediately afterward, status during polling, and output URL at completion. That record lets a reviewer distinguish a failed request, an unfinished request and a completed clip without relying on a screenshot or memory.

Worked example: a controlled first candidate

Hypothetical seller case: one approved front-facing fashion image is available, no final pose is required, and the seller needs an internal motion review. The seller leaves prompt empty, omits end_image, selects 5 seconds at 720p, submits the request with the documented model and nested inputs, and records the returned prediction ID.

This is planning guidance, not an executed FASHN request. If timing is unreadable, the next controlled change is 10 seconds at 720p, raising the listed cost from 3 to 6 credits. If timing is acceptable but garment detail is insufficient, retain 5 seconds and test 1080p, which also raises the listed cost from 3 to 6 credits. The two changes should not be combined until the separate review questions are answered.

Where CartVisual fits

CartVisual's product-video workflow is the relevant preparation route for creating or reviewing a product-video candidate. It does not implement FASHN's API or automatic polling.

See the current CartVisual workflow and its limitations

Source and evidence scope

Prepared by the CartVisual editorial team with AI assistance. Competitor behavior is documented rather than independently executed for this guide.

FASHN: Image to Video

Official source snapshot checked .

Vendor facts are limited to the supplied official FASHN Image to Video documentation, checked 2026-11-17. The source labels the model Experimental. Configuration recommendations and the worked case are original editorial guidance; no FASHN request or result was executed.

Related decisions

Browse ecommerce creative workflows