Skip to content

Qwen Image Edit / Image editing

Qwen edit to MiniMax H3 video: approve the still first

Separate scene design from motion. First use Qwen Image Edit to prepare a still, then use the approved result as the source for H3 image-to-video. This gives you a checkpoint where composition and product accuracy can be fixed before spending time on animation.

By JollyAIReviewed 2 min read

Decide which stage owns each change

Use the image edit for background, lighting or a deliberate visual treatment. Use the video prompt for movement and sound. Asking H3 to repair a weak composite while also animating it makes the result harder to diagnose.

Keep the original photograph and the edited still. If a later video changes the subject, these files help identify whether the problem began in the edit or in the animation.

Two-stage workflow

  1. Edit the source in Qwen with a narrowly defined change.
  2. Compare the still with the original and approve its outline, details and lighting.
  3. Download that still and upload it to H3 Image to Video.
  4. Request a restrained camera or environmental movement without redesigning the scene.

Example paired briefs

PROMPT RECIPE / ADAPT TO YOUR SOURCE

Qwen edit: preserve the exact green bottle and its cap; replace the background with a dark studio wall and a matte pedestal, matching the existing light. H3 motion: preserve the approved image; make a small slow camera push-in while a soft highlight moves across the bottle; quiet studio ambience.

Review both outputs separately

The Qwen stage creates an image. A before-and-after wipe of that image is a presentation animation, not evidence that Qwen generated motion. The H3 stage creates the video. Label demonstrations accordingly and inspect the whole motion clip for drift. If the video fails, return to the approved still instead of repeating the image edit unnecessarily.

REVIEWED JOLLYAI IMPLEMENTATION

The workflow behind the guide

One to three images + edit brief → Qwen rapid SFW checkpoint and image conditioning → sampling → decoded edited still.

Inputs, presets & access
Image output. Use your current account allowance; a subsequent video request is a separate generation.
Implementation note
The reviewed JollyAI edit route loads Qwen-Rapid-AIO-SFW-v23 and accepts one to three images. It returns an edited still. It is a configured derivative workflow, not a claim that JollyAI serves the unmodified upstream Qwen-Image-Edit-2511 checkpoint. Current account allowances appear in the playground.

Configuration reviewed September 20, 2026. Your current playground controls and plans page determine availability. Prompt recipes above are suggestions, not guarantees or claimed test results.

Open Image editing ↗

Sources and useful next steps

The implementation notes describe JollyAI's reviewed local integration. Upstream capabilities and licences may differ from the settings available here.

Found an outdated setting or a reproducible problem? Contact JollyAI with the workflow and job reference. Keep account tokens and private input files out of public reports.

KEEP BUILDING

Continue the workflow.