← Back to blog
MaxVideoAI · Creative workflows

Build an Animated Café Menu with Editable Prices Using MaxVideoAI, FLUX 3 and Codex CLI

A documented workflow for designing two text-free anchor images, requesting a quoted FLUX 3 transition through MaxVideoAI in Codex CLI, adding editable menu copy in an external editor and checking the loop on its destination display.

By MaxVideoAIPublished September 25, 2026
Illustrated café display with a steaming cup beside a blank menu area and a separate floating overlay.
Illustration. Keep the atmosphere behind the information.

Design both anchors as one composition

Example prompt
Create two matching, text-free background images for a landscape café menu display.

Use an identical locked camera, crop and perspective in both images. Reserve the entire left half as a calm, low-detail warm-cream area for menu typography that will be added later in an editing application.

On the right, place the same terracotta coffee cup and slice of toast in exactly the same positions. Use a warm cream, terracotta, espresso-brown and amber palette with stable, intimate lighting.

Start image: show one short, fine curl of steam above the cup.

End image: preserve every object, shadow and compositional boundary, changing only the steam so that it has drifted slightly upward and to the right. Keep the steam entirely outside the reserved menu area.

Do not include dish names, prices, numbers, logos, labels, signage, watermarks or interface elements. Do not add, remove or reposition props between the two images.

Use this brief in an external graphics or image-creation tool. The fixed camera, stable objects and consistent lighting leave the eventual transition only one modest change to reconcile: the steam. Keeping the left side quiet also protects the readability of menu copy added later.

Prepare two images that differ only where motion is needed

  1. 01

    Match the destination display

    Build both images at the same aspect ratio as the café screen. Use one shared composition guide so the crop, counter line, cup, toast and reserved left-hand area remain aligned.

  2. 02

    Create the start image, then derive the end image

    Finish the text-free start composition in an external graphics tool. Duplicate it for the end image instead of rebuilding the scene, then make only the planned adjustment to the steam.

  3. 03

    Remove critical copy from both images

    Export backgrounds without names, prices, logos or readable signage. FLUX 3 guidance recommends keeping critical text out of generated video because readable signs, small interfaces and micro-labels can lose fidelity.

  4. 04

    Compare the anchors directly

    Overlay or alternate the two images before export. Correct any unintended shift in framing, object placement, palette or lighting; the steam should be the only meaningful difference.

  5. 05

    Check the required MaxVideoAI inputs

    Before preparing the generation request, review the MaxVideoAI FLUX 3 model page at https://maxvideoai.com/models/flux-3. Its documented first-and-last-frame website workflow requires one prompt, one Start image and one End image. This website schema describes the required inputs, but it does not establish that the same mode is available through the connected Codex route; that check comes before requesting a paid generation.

Change the steam, not the layout

Two matching text-free café backgrounds differ only in the small steam curl above the cup.
Illustration. Start above, end below: preserve the layout and give the steam only a modest change to make.

Keep the motion away from the menu copy

Example prompt
Create a restrained, physically plausible transition from the supplied start image to the supplied end image. Keep the camera locked. Preserve the cup, toast, counter, framing, warm color palette and lighting. Animate only a slow, delicate drift of steam above the cup, following the change shown between the two anchors. Keep the blank left-hand menu area calm, empty and undisturbed. Do not introduce lettering, prices, logos, signage, new props or additional camera movement.

With the matched anchors ready, give the transition only one modest job. Keeping movement around the cup rather than behind the reserved menu area protects legibility and reduces the changes the bridge must reconcile. Choose duration, resolution and other operational settings only after Codex reports what the connected catalog supports and includes them in the exact quote.

Ask Codex to verify the route before generating

Example prompt
Help me plan a MaxVideoAI FLUX 3 first-and-last-frame video generation.

Start image: [ATTACH OR REFERENCE THE START IMAGE]
End image: [ATTACH OR REFERENCE THE END IMAGE]
Motion prompt: [PASTE THE RESTRAINED-MOTION PROMPT]

First inspect the connected catalog and confirm whether the requested FLUX 3 first-and-last-frame workflow is available. Preserve both supplied images as required start and end anchors. Report the settings supported by this route and present the exact quote before any paid generation.

Do not generate or incur a charge until I explicitly approve that quote. If the route, the start-image input or the end-image input is unavailable, stop and explain what is missing. Do not substitute text-to-video, single-image video or another model or mode.

Replace all three bracketed references before sending this in the Codex conversation that has the MaxVideoAI connection. The request separates catalog inspection from paid generation, so an unavailable route produces a clear stop decision rather than a silent change of workflow.

Approve the route and quote, not an assumed result

  1. 01

    Use the documented Codex setup flow

    Open the reviewed MaxVideoAI Codex integration guide at https://maxvideoai.com/integrations/codex, follow its setup flow, start a new Codex conversation and complete OAuth when prompted. Codex CLI hosts the request; MaxVideoAI is the boundary that performs the requested generation.

  2. 02

    Treat compatibility evidence as version-specific

    The supplied integration excerpt reports contract checks for Codex CLI 0.150.0-alpha.8 on August 27, 2026. That report does not establish compatibility with a current or different CLI release, Codex Desktop or ChatGPT, and it does not independently verify this particular FLUX 3 request.

  3. 03

    Confirm the requested input structure

    Before considering the quote, check that the proposed job retains the motion prompt, the start image and the end image. The reviewed MaxVideoAI FLUX 3 page at https://maxvideoai.com/models/flux-3 documents that three-part structure for its first-and-last-frame workflow. The distributor endpoint corroborates the same structure, but neither source proves that the connected Codex route is currently available or that generation will succeed.

  4. 04

    Review the exact quote before approving

    Check that the quote describes the requested FLUX 3 first-and-last-frame route and only settings you intentionally selected from the connected catalog. Approve paid generation explicitly only if the route, inputs, settings and cost are acceptable. Otherwise, stop or revise the request.

  5. 05

    Continue only with a returned candidate clip

    After approval, MaxVideoAI may return a candidate background bridge if generation succeeds. Treat it as material to inspect in the external editor, not as a guaranteed seamless loop. If the route is unavailable, authorization fails, the quote is rejected or no clip is returned, stop before the menu-assembly stage.

From two images to an editable menu

Prompt + two images
Submit brief and anchors
Codex CLI
Check route; request quote
Exact quote
Human approves generation
MaxVideoAI · FLUX 3
Return clip if successful
Candidate background clip
Import; add editable copy
External editing project
Codex CLI hosts the request; MaxVideoAI generates the background; editing and display checks remain external.

Assemble the menu so tomorrow’s prices stay editable

  1. 01

    Build above the candidate background

    Continue only after MaxVideoAI returns a candidate clip through the quoted, approved Codex workflow. If no clip was returned, stop and revisit the documented integration route at https://maxvideoai.com/integrations/codex. Place the candidate clip on the lowest track in your external editor. Add the dish names and prices as separately authored graphics above it—for example, Coffee and Toast on one layer, with $4 and $6 on another. Keeping commercial copy outside the generated imagery makes later revisions practical.

  2. 02

    Protect the text from competing motion

    Position the copy inside the quiet area reserved by the anchor images. If steam, lighting changes or background detail weaken the contrast, add a stable cream backing panel between the clip and the menu graphics. This preserves readability without flattening the atmosphere across the entire display.

  3. 03

    Retain the layered editing project

    Save the project with the background, backing panel, dish names and prices separated. When Coffee changes from $4 to $5, you can reuse a still-suitable background, although typography, spacing and the final export may still need adjustment.

  4. 04

    Handle sound outside the generated clip

    Mute or remove any unwanted audio in the editor. A menu display rarely benefits from an accidental soundtrack, and silent export avoids depending on the destination player’s volume settings.

  5. 05

    Construct a proposed repeat

    Place copies of the background clip end to end and inspect the join. If the cut feels abrupt, try trimming the background or testing a short overlap on the background only. Watch that overlap carefully: blending can double the steam, cup or other objects instead of hiding the transition.

Keep names and prices above the background

An exploded menu separates the café background, cream backing panel, dish names and prices into four aligned layers.
Illustration. Build the menu in an external editor, with names and prices separate from the background.

Decide whether the loop works on the actual display

  1. 01

    Inspect the end-to-start join

    A clip can connect its supplied start and end images successfully yet still jump when its final frame repeats back to its first. Concentrate on the cup, toast, lighting and steam at that separate end-to-start boundary.

  2. 02

    Preview the proposed repeat continuously

    In Adobe Premiere, mark the repeat region with In and Out points, enable Loop playback in the monitor and start playback with Space or L. Let it run several times while you evaluate the transition timing; a single pass can make a small jump easy to miss.

  3. 03

    Read the menu from the customer’s position

    Export a test and play it on the destination screen. From the expected viewing distance, confirm that Coffee, Toast and their prices remain legible, with enough contrast and spacing for quick scanning.

  4. 04

    Check motion, playback and sound together

    Watch several full repeats for a visible seam, frozen moment or pause introduced by the destination player. Confirm that the steam never competes with the prices, and listen for audio that should have been removed or muted.

  5. 05

    Revise or export

    If the join draws attention, adjust the trim, overlap or background choice and test again. If the copy stays readable, the motion remains secondary, the repeat is unobtrusive and playback is silent as intended, export the delivery version. Loop preview helps reveal problems, but only the destination screen and player show how the menu will actually behave.

Look closely at the repeat

An illustrated café menu sits above two enlarged steam details showing a possible jump at the repeat boundary.
Illustration. Compare the end on the left with the restart on the right; a small steam jump can reveal the join.
Sources · 5
  1. [1] Create AI videos with Codex

    MaxVideoAI · 2026-09-25

  2. [2] FLUX 3 Video

    MaxVideoAI · 2026-09-25

  3. [3] Flux 3 First Last Frame to Video API

    fal · 2026-09-25

  4. [4] Package Your Plugin

    OpenAI · 2026-09-25

  5. [5] Play Back in a Continuous Loop

    Adobe · 2026-09-25

About MaxVideoAI

MaxVideoAI

MaxVideoAI publishes practical guides, model comparisons, and production examples. Its editorial standards distinguish provider specifications, observed measurements, and editorial judgments.

Related reading

More workflow notes and engine breakdowns curated for you.

Oct 25, 2026

How to Change Your Video Background with AI | Wan 3 Prime

Take one talking clip from a home office to a spacecraft or an underwater observatory. Follow the source selection, settings and exact prompts from our Wan 3 Prime Modify tutorial.

Read article

Sep 28, 2026

Turn a Voiceover into a Focused Visual Capsule with LTX 2.5 Pro

Learn how to translate one narration cue into a clear visual idea, prepare the required audio and optional opening image, review the candidate clip, and finish the capsule with titles and captions in an external editor.

Read article

Sep 27, 2026

How We Made an AI Podcast with Codex and MaxVideoAI

Six generated takes, one 86-second interview. Inside the prompts, character references and Premiere Pro edits behind our fictional AI podcast.

Read article