Do not paste a still-image recipe into a video job
A merchant has an approved Midjourney still of a coffee machine, created with image styling and reference parameters. They want a short camera push. The still image can become a starting frame, but its image-generation parameters do not automatically become video controls. Midjourney's current Video documentation explicitly removes original image parameters when animating and defines a separate video parameter set. The correct migration preserves the approved image and rewrites the motion instruction.
Map the input contract
| Input category | Still-image role | Video treatment |
|---|---|---|
| Image-generation parameters, including style-reference settings | Shape the original still | Not transferred as video-generation controls |
| Style Reference | Image style guidance | Unsupported as video reference input |
| Edit Model Reference | Image editing reference | Unsupported as video reference input |
| Image Prompt | Image prompt reference | Unsupported as video reference input |
| Starting image | Finished still or uploaded source | Use the entry appropriate to gallery, web upload or Discord source |
| Video motion text | Not a still-image parameter | Describes movement and scene behavior |
Do not interpret an unsupported reference type as evidence that the starting frame itself is invalid.
Choose video controls for their actual jobs
| Control | Purpose | Not a guarantee of |
|---|---|---|
--motion low / --motion high | Requested motion level | Exact camera path or unchanged product geometry |
--raw | Video-specific raw setting | Physical product authenticity |
--loop | Loop-oriented start/end behavior | Perfectly seamless intermediate movement |
--end | Optional end-frame input | Correct motion between the frames |
--bs 1, --bs 2, --bs 4 | Number of video candidates | Improved accuracy or a longer clip |
--video in Discord with own-image URL | Submission flag for the documented own-image Discord route | An additional image-style control or general video quality setting |
The --video submission syntax is separate from the provider's enumerated video-generation controls. Do not claim it is a transferable still-image parameter.
Hypothetical migration: coffee-machine teaser
First save STILL-04, the approved machine image showing the correct dial, handle and printed brand. Use the entry that matches where this source actually lives:
- Midjourney gallery image on the website: open the image and use its Animate Image controls. Animate Manually lets the operator adjust the motion text before submission.
- Your own external image on the website: upload or select it through the Imagine bar's image panel, then place it in the Animate section.
- Your own image through Discord: use the documented image-URL submission with
--video.
The official Video guide describes these separate entry routes. After the appropriate entry, prepare the video stage: write a fresh motion brief, select supported video controls, and inspect the returned candidate. For this fictional teaser, a proposed motion brief is “Slow forward camera movement; machine stays in place; no new buttons or operating sequence.” Style Reference and Image Prompt are not alternate video input slots. These entry routes are alternatives, not flags to combine.
Review the proposed opening frame, dial geometry, lettering, background continuity and final frame. A loop request does not prove a seamless seam, and an end frame does not prove physically accurate movement. Midjourney currently describes five-second initial clips and four-second extensions, up to four extensions; treat those as documentation-specific current limits, not permanent cross-version specifications. No video was generated here.
Close the parameter migration
Record the still's original recipe for provenance, then store the separate starting image, video text, supported video settings, chosen route and actual result identity. For another platform's source-selection problem see Product Studio image provenance; for broader first-frame distinctions see first frame versus reference video. Neither provides a Midjourney parameter workaround. Reject any clip inventing a product function or changing the physical SKU.