Mai Image 2.5 Edit is not video

Mai Image 2.5 Edit is a 5-credit Microsoft stills editor. It is not a video row. Lock the pixels, then animate. Mai Image 2.5 Edit is Microsoft, content type image, categories edit-image and image-to-image, 5 credits, requires an image, audio false, no duration list, no 4K badge on the published card. The catalog line: MAI-Image-2.5 image editing — pixel-level, cleanup, backgrounds, text. Features listed: edit_image, image_to_image. This page is the category error: Mai 2.5 Edit is not motion, not a talking presenter, not a Microsoft video SKU we did not publish. If the launcher shows seconds, you left this slug.

There is no second Mai video card in the Versely published list to "upgrade" into. Do not invent one. After the still is locked, pick a real I2V row.

Microsoft's stills editor

Read the fields.

• ctype: image
• cats: edit-image, image-to-image
• credits: 5
• requires_image: true
• durs: empty
• audio: false
• max output: unset — do not stamp 4K because other editors list it
• aspects: auto, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3
• max_reference: unset on this card
• provider: Microsoft

Requires an image is the whole product. You cannot describe a living room from scratch here. Upload the locked still. Write the change. Pixel-level, cleanup, backgrounds, text — those are stills nouns. Cleanup is dust, wires, a stain, a hand you do not have a release for. Backgrounds is a sweep or a swap you can point at. Text is in-image lettering you are correcting or replacing, not captions burned onto a clip.

Five credits sits between cheap instruction editors and heavy family edits. Pruna Image Edit is 1 credit and 4K. Qwen Image Edit 2511 is 3 credits. Wan 2.7 Pro Edit is 8 credits with up to four references. Pay 5 when the brief is a Microsoft MAI stills edit: cleanup, background, in-image text on a file you already have. Do not pay 5 because the name looks like a flagship video model.

Audio false is not "silent video." There is no video. False is a stored flag on an image card. Do not prompt a voiceover.

Match the source aspect. Auto is on the list; named ratios are on the list. A 9:16 edit of a 1:1 product plate is a crop. Do the crop on purpose or keep the plate.

Why it gets treated as video

Three habits.

1. The name sounds like a versioned media stack. "2.5 Edit" reads like a clip editor if you live in Premiere. On this catalog it is pixels. Teams then say "Mai video is weak." They never left stills.

2. The prompt is written like a shot list. "Slow push in, 6 seconds, hold the label." Mai will try to draw a push-in. You will get a still that looks like a frame from a push-in. The label may still be wrong. You will have spent 5 credits to illustrate a camera.

3. Microsoft is filed next to Copilot demos in someone's head. A demo that shows a moving thumbnail is not this slug. This slug requires an image and returns an image. Trust the content type over the last keynote you remember.

The correct sequence:

1. Get a still you would print — camera or T2I. Text to image if you still need a first plate.
2. Edit on Mai until cleanup, background, or in-image text is right. One change per run if the last run drifted.
3. Send the approved file to image-to-video only if the deliverable needs time. Hailuo 2.3 Fast is 6s / 10s at 28 credits. Omni Video is 4–10s at 9 credits. Veo 3.1 is 4 / 6 / 8 seconds, 4K, 40 credits, native audio, when the shot is Veo.

Keep the still. The next variant should start from this file, not from a new paragraph that almost describes it.

Pixel jobs this row owns

Mai is for:

• Cleanup you can circle on a screenshot: the extra cup, the wrinkle that is not the product, the exit sign.
• Background replacement when the product must not drift.
• In-image text correction when the string is already in the pixels and you can quote the replacement.

Mai is not for:

• First generates. No source, no job.
• Duration. Empty durs.
• A talking head. That is lipsync or T2V, not an image editor.
• Bulk 1-credit retouch you could do on Pruna. Five is the wrong sticker for forty speckles.
• Dense packaging type as a layout generate. That is Seedream 5.0 Pro on a blank page, or a Qwen type editor if you are patching a file.

Versus other stills editors, briefly: Qwen Image Edit lists 4K at 3 credits and is the cheap type-fix. 2511 is the dated instruction SKU. Flux Kontext is 4 credits, 4K, Flux coherence. Mai is Microsoft's 5-credit cleanup / background / text editor. Pick it when that sentence is the brief, not because Microsoft is the office suite.

After the edit

If the deliverable is a still, you are done when you would print it. Do not "just see it move" as a quality check. Motion will invent time and, with it, new dirt, new type, new fingers.

If the deliverable is a clip, I2V on the cleaned file. The AI video generator is the Versely version of "run the motion model." It will not run Mai as a clip.

Captions happen on the video. Add captions to video. In-image text you already corrected on Mai is not a caption track. Do not recaption a label you legalled in pixels unless the clip needs spoken-word captions for silent autoplay.

Article 50 can apply to a realistic still of a real person. Cleanup does not exempt it. Read Article 50. New York's synthetic performer rule (N.Y. GBL § 396-b, in force 9 June 2026) is an ad-disclosure statute for synthetic human performers, not a reason to treat Mai as video. If the still is a generated person in an ad reaching NY consumers, that is a disclosure question. It is still a still.

The test: is the source a PNG you would print except for cleanup, background, or a quoted string? Mai, 5 credits. Is the source a clip? A video-edit or I2V slug. Is the source a sentence and no file? T2I. "Mai 2.5" without "edit" and "image" is how a stills editor gets blamed for not rolling a camera.

Confirm mai-image-2-5-edit, then sign up when the job is a 5-credit stills edit, not a Microsoft-shaped clip. Mai stays an image row.

FAQ

Is there a Mai video model on Versely?

Not in the published catalog this page is written against. Do not invent a T2V cousin. After the still, pick a named video slug.

Can I skip the source image if the prompt is detailed?

No. requires_image is true. A detailed prompt without a file is a text-to-image job on a different card.

Does audio false mean I get a silent MP4?

No. You get a still. Audio false is not a 0 kbps soundtrack. There is no timeline.