Agent Iris Creative Studio: Amazon Images and Video

Amazon Agent Iris is a creative studio for Amazon product images and video, operated through your AI client. It combines generation tools with built-in image and video Skills that guide the brief, destination-specific requirements and review. Results can include signed URLs for download and further work. Iris is a Kuudo product, separate from Amazon Creative Studio.

You can plan a product image gallery, create and review main or alternate images, turn product references into video shots, and prepare media for the appropriate Amazon placement. Iris helps check the work; you still review the output and Amazon makes the final acceptance decision.

This page covers what Iris does, the rules it applies, what you can ask it for, which model key to bring, and how listing media moves through the workflow. Choose Managed MCP setup for a service operated by Kuudo, or Private Cloud for infrastructure in your account. You connect either path from the AI client you already use.

The creative methods inside Iris

Iris carries two Skills as MCP resources. They are instructions and references for the assistant to read while working, distinct from the public Kuudo Skills installed in a client. They are not separate MCP servers or an automatic guarantee of compliance.

After connecting Iris, ask your assistant to read the matching Skill before planning or generating. Clients that expose MCP resources can read the URI directly; clients that expose tools can call read_resource with that URI. start_here provides a server-side workflow guide, and list_resources helps discover available resources. The Skills route the assistant to the reference material relevant to the task.

For example, say: “Read the Iris Amazon product image Skill, then plan a gallery for this product from my reference photos. Tell me which facts or views are missing before generating.” For video: “Read the Iris Amazon video Skill, then plan a product-detail-page video using these photos. Show the beat sheet and preflight before any billed generation.”

Plan the product image gallery

Start with the actual product, its category, marketplace and intended image slot. Supply clear, authorized reference photos, verified dimensions and included contents. The Skill separates main-image requirements from the creative choices available for alternate images, and helps choose images that answer different buyer questions:

• Main image: identify the product clearly and review the background, framing and included items against the applicable requirements.
• Feature and detail images: show construction, controls, materials or texture from a reference that actually shows them.
• Infographics: explain verified dimensions, contents or compatibility in readable text without unsupported claims.
• Lifestyle and instructional images: show a plausible use, fit or setup when the category and placement allow it.

Ask Iris for a slot-by-slot brief: the buyer question, proposed image, reference source and facts to verify. Gallery order is a merchandising recommendation, not Amazon policy. Do not ask the model to invent an unseen angle, label or product part; provide another reference when the detail matters. Review identity, colors, quantities, labels and claims in every result at the intended display size.

For a main image, a white-looking background is not proof of exact white pixels. The Skill calls for deterministic downstream background normalization and pixel sampling where exact white is required; Iris does not perform that normalization. Consult the current Amazon rules and the applicable category before submitting.

Plan and direct Amazon video

First name the destination: a product-detail-page video, Sponsored Products video or another ad product. Their requirements differ, so check the current marketplace and upload surface before treating a format or duration as eligible. Use real product references and decide what the viewer should understand: an overview, a feature demonstration, an unboxing, setup, troubleshooting or a product-focused brand story.

Ask Iris for a beat sheet covering the visible action, approximate timing, framing, reference and review criterion. A fresh clip can contain several shots; separate clips are useful when a shot needs its own reference or independent revision. A consistent brief helps continuity but does not guarantee it. Check labels, parts, proportions, motion, sound and transitions against the real product.

Iris can generate a fresh 3–10-second clip and edit or extend an eligible Iris-owned video. Extensions require an owned handle and retained lineage, and can reach up to 40 seconds total. For a longer finished video, plan an external edit with shot order, trims, transitions, audio and intended runtime. Iris does not stitch clips, remove audio, export the final video or upload it to Seller Central. Review the exact prompt, settings and available estimate with validate_video_plan before a billed submission; an unavailable estimate is not free.

Creative work beside your Amazon accounts

• Sellers: the Selling Partner MCP supplies live catalog context and handles supported, approved listing-image changes and ingestion verification.
• Vendors: use brand references and authorized account context alongside Vendor MCP operations. Confirm the media-delivery route for the account and asset; vendor account access does not establish the same publishing path as seller listing images.
• Advertisers: Iris creates placement-specific assets, while the Amazon Ads MCP registers assets and performs supported creative and campaign operations in the authorized account.

Your client coordinates these connections. Iris does not replace their permissions or approval policies, and it is not a visual canvas or timeline editor. Video assembly, final export and Seller Central video upload remain external.

Pair Iris with Amazon Listing Optimizer

The Amazon Listing Optimizer Skill and Iris can work together in an AI client that has the Skill and both the Selling Partner MCP and Iris connected. The optimizer audits the authorized seller listing and proposes listing fixes; Iris helps create and review the product images and video the team needs. Your client coordinates the handoff using the information and approvals you provide.

A connected workflow for one SKU can run like this:

1. Find the gaps. Ask the optimizer to inspect the listing, including existing images, copy, attributes and reported issues. Select the gaps worth addressing; do not infer a product feature from a missing image.
2. Prepare the creative brief. Supply authorized product photos and verified facts. Ask Iris to read the product image Skill, propose the missing main or alternate images, and explain which references or facts are still needed. For a video opportunity, ask Iris to read the video Skill and plan for the intended placement.
3. Create and review. Review the briefs, generation settings and expected or available costs before billed work. Inspect generated images and video against the actual product, and verify the applicable Amazon requirements. Request changes where details, labels or claims are wrong.
4. Preview and approve listing changes. For a supported seller image update, provide or explicitly approve the reachable, unexpired Iris image URL as the source for validation. Ask the optimizer to preview the exact listing patch; the Skill requires your explicit confirmation before a write. Submit the approved change through the Selling Partner MCP, then check Amazon's response and ingestion.
5. Finish video separately. Iris returns clips for review and can edit or extend eligible Iris-owned video. Give the approved clips and finishing notes to an editor for assembly, final export, applicable metadata and Seller Central upload; the listing-image patch does not upload a video.

Your client can automate the coordination of these steps across a selected set of listings, with the review and approvals described above. The handoff depends on the client passing the approved assets between the connections; it is not a built-in Iris-to-optimizer integration. The optimizer also has its own separate client-side image-generation workflow, which requires the provider credentials and storage described in that Skill. Keep the account, scope, generation costs and required approvals clear for each action, and review the resulting assets and listing changes.

Try this in your connected client:

Use Amazon Listing Optimizer to audit [SKU] in [marketplace].
Identify missing or weak listing images and any product facts that need confirmation.
Read the Iris product image Skill and propose a gallery using my supplied photos.
Also read the Iris video Skill and outline a product-detail-page video.
Show the plans and expected or available costs before generating.
After I review the assets, preview any listing image patch and wait for my approval.

See the Listing Optimizer guide for the audit and approval workflow and the Iris connection guide for the separate creative connection.

What it does

The server handles images and video through natural-language requests:

• Generate images — create a main, infographic, lifestyle, detail, or fashion image for review.
• Edit images — fix a compliance problem or change one element while preserving product identity.
• Audit images — review an image against the applicable rule set and propose fixes; verify the findings before submission.
• Generate video — create a 3–10-second clip from text, a product image, or first and last frames. A clip can contain multiple shots.
• Edit or extend video — change an owned Iris video with a focused instruction, or add 3–10 seconds per request up to 40 seconds total. Valid retained lineage is required.

The same server handles creation and review, so you can build the listing's still-image stack, plan its video, and regenerate a shot that fails product-fidelity review without leaving the conversation.

Example prompts

You drive Iris in plain language from your AI client — no tool names, fields, or APIs to remember. Replace bracketed values like [ASIN] (Amazon Standard Identification Number), [SKU], or [brand] with your own.

Find problems across the catalog

Surface what needs attention before you decide where to spend effort.

• "Scan my US listings and show me which ones have image problems."
• "Which of my live listings have empty image slots — fewer images than Amazon allows?"
• "Find my listings that are suppressed and tell me why."
• "Show me listings that only have a single image."
• "Give me a health check on [parent ASIN] and all its variations."

Diagnose one listing

Zoom in on a specific SKU to understand its current state.

• "Pull up [SKU] and tell me what's wrong with it."
• "Review the listing for [ASIN] — title, bullets, description, and images."
• "Why isn't [SKU] showing in search? What does it need to come back online?"
• "How many images does [ASIN] have, and what's missing?"

Fix listing copy

Clean up or rewrite text, grounded in the product's real attributes.

• "Rewrite the bullet points for [SKU] — it only has one and it's weak."
• "The description for [SKU] has copy-pasted text from another product. Clean it up."
• "Improve the bullets for [ASIN] using only what's true from the listing — don't invent features."
• "Make the title for [SKU] clearer and keep the brand out of the body copy."

Generate product imagery

Create new images from your real product. Lead with the scene you want.

Lifestyle / in-use shots

• "Create a lifestyle photo of [product] styled on a cream sofa in a bright, neutral living room."
• "Show a person using [product] in a warm, natural-light kitchen scene."
• "Make a cozy bedroom scene with [product] as the focal point."

Hero / product shots

• "Generate a clean studio shot of [product] on a pure white background."
• "Zoom in on the [product] so it fills more of the frame and reads as the hero image."

Infographics (added text is allowed on these auxiliary images)

• "Build a size and dimension graphic for [SKU] using the real measurements from the listing."
• "Create a feature-callout image for [product] highlighting its material, stitching, and zipper."
• "Make a care-instructions card for [product] — machine washable, etc."
• "Create a 'set of 2, covers only — no insert' graphic so buyers know what they're getting."
• "Make a color-range image showing all the available colors for this variation family."

Generate listing-video shots

Start with the Amazon video type and the product references Iris should preserve. Product overview, unboxing, how-to, setup, troubleshooting, and product-focused brand-story videos are supported planning patterns.

Amazon Ads images and video use the same generation surface, but not the listing rule set. Name the ad product and placement before generation so Iris can apply the destination-specific brief; the Amazon Ads MCP handles the connected creative and campaign operations.

• "Create a 16:9 product-overview shot of [SKU] from this product photo. Keep the label, color, controls, materials, and included parts exact."
• "Animate this hoodie reference into a slow studio reveal. Keep the zipper open and preserve the sherpa lining, drawcords, zipper, and hem tab."
• "Plan a product overview for [ASIN], then preflight the first 8-second feature clip at 720p before generating it."
• "Create a vertical 9:16 shot of [product] for the listing, using this product photo as the visual reference."
• "Extend the last Iris video with the next seven seconds of the same continuous demonstration."
• "Check this generated clip frame by frame for label, color, control, included-part, and proportion drift."

Edit or extend an Iris video

Use the owned Iris video handle from a previous result. Ask for a narrow change, such as changing the lighting, or describe what happens next in an extension. Arbitrary uploaded-video editing is not exposed.

• "Change the lighting in the last Iris video to soft morning light. Keep everything else the same."
• "Continue this Iris sequence for five more seconds, then review the transition and product details."

For edits, omit duration to preserve source timing as the goal, then check the returned result. Extensions return the accumulated video: do not append that result to the source again. Review continuity, labels, packaging, proportions, and audio after either action. Preservation is not guaranteed.

An edit or extension is a new paid creation. Preflight the same arguments you intend to submit and approve the request before generation. A failed fidelity review does not authorize an automatic paid retry.

Edit and iterate on an image

Refine an image you've already generated, in plain language.

• "Zoom in slightly on the pillow in this image."
• "Remove the books in the background — keep everything else the same."
• "Same scene, but brighter and with more natural light."
• "Make the product larger and more centered."
• "Try this again on a grey sofa instead of cream."

Check compliance before publishing

Validate against Amazon's image rules before anything goes live.

• "Check whether this image meets Amazon's compliance requirements."
• "Is text allowed on this image for the slot it's going into?"
• "Validate the file format, resolution, and color space on this image."
• "Run the policy check on all the images I've generated so far."

Audit an image you already have

Iris audits an uploaded photo the same way it audits a generated one. Ask for an upload link, send the file, then run the check.

• "Give me an upload link for a photo I already have."
• "Audit the image I just uploaded against the main-image rules for [SKU]."
• "Fix the background on the uploaded photo and keep everything else the same."

Uploads go through create_upload_url: a short-lived signed URL that accepts JPEG, PNG, WebP, or GIF files up to 40 MB, with no API key on the upload itself. A stored image may receive a signed download URL, which can be used for a listing update while it remains reachable; confirm its expiry and verify ingestion. Files made in another tool, including a design tool's export, take this path.

Publish to the listing

Write approved changes directly to the live listing.

• "Add this image to [SKU] as the first alternate image."
• "Put the size graphic into the next open slot on [SKU]."
• "Patch [SKU] with this image."
• "Apply the rewritten bullets and description to [SKU]."

Confirm it actually went live

Verify the change was accepted and the image was ingested by Amazon.

• "Check whether the new image on [SKU] ingested cleanly."
• "Did my last change to [SKU] go through, or are there any errors?"
• "Re-check [SKU] and confirm the gallery now shows the images I added."

Reuse imagery within a variation family

Fill gaps by borrowing the right images from sibling SKUs — safely.

• "This color is missing alternate images. Which images from its sibling colors can I reuse without misrepresenting it?"
• "Show me the best-stocked sibling of [SKU] and which of its images are color-neutral enough to copy."
• "Fill the empty slots on [SKU] with the generic infographics from the family, but not the color-specific shots."

Work at scale

Move from one listing to many.

• "Make a list of every listing missing a main image so I can prioritize them."
• "Which listings would benefit most from a lifestyle image? Rank them."
• "Walk me through fixing the thinnest listings one at a time."

Most real sessions chain these together. An image workflow might run: "Scan my listings for image gaps" → "Let's start with [SKU]" → "Create a lifestyle photo on a cream sofa" → "Check it for compliance" → "Add it to the listing" → "Confirm it ingested." A video workflow might run: "Plan a product overview for [ASIN]" → "Use these two references" → "Generate the hero shot" → "Review product fidelity" → "Generate the remaining shots and give my editor the order and finishing notes." Treat either as a conversation, not a set of one-off commands.

The audit verdict

When you ask for an audit, the skill returns a structured verdict rather than a vague opinion:

Status: Compliant | Needs changes | Reject
Scope: [main image / alternate / fashion / multipack / ...]
Blocking issues:
- [issue and the rule category it violates]
Non-blocking suggestions:
- [quality or conversion improvement]
Fix:
- [the exact edit instruction or replacement prompt]
References checked:
- [which rule references were applied]

Use the verdict to organize review and fixes by rule category. For exact-white requirements, verify the background through pixel sampling rather than relying on visual appearance or an unverified verdict.

Image requirements and review

The amazon-product-image Skill carries a distilled rule set from Amazon Seller Central policy to guide generation and audits. Check current marketplace and category requirements and review the resulting asset before submission.

Main image — hard requirements

Technical file requirements

All-image content rules

These apply to every image in the stack, not just the main:

• No reviews, stars, ratings, prices, deals, coupons, or free-shipping claims.
• No seller info, email, copyright marks, or watermarks.
• No Amazon, Prime, Alexa, "Choice," or "Best Seller" branding or lookalikes.
• No promotional, warranty, certification, or unsubstantiated safety/health/regulatory claims.
• Every image must match the product title, ASIN, variant, color, quantity, and scale.

Special cases

Suppression issues and fixes

The skill maps the common suppression and rejection reasons to a concrete fix:

Fashion, apparel, and footwear

• Adult apparel on-model and standing; kids, babies, accessories, and multipacks off-model (flat laydown).
• Framing by garment: full-length for dresses and suits, top-body for shirts, waist-down for pants and skirts.
• Footwear main image: single shoe, left foot, 45°.
• Off-model laydowns: white surface, steamed, loose threads removed, square framing.
• A 13-point fashion validation checklist covers parent/child coverage, naming (ASIN.MAIN.jpg, ASIN.PT01.jpg, ASIN.FL01.jpg), the image set, and mobile readability.

The image stack

Most listings stop at the main image. The skill helps you build the full stack that actually converts:

Image and video providers

One server exposes Google Gemini and OpenAI image engines alongside Google Gemini Omni video. For Google generation on Managed MCP, normally get a Gemini API key and provide it as described in the Iris connection guide. In Private Cloud, bring your own provider key; Kuudo does not sell creative credits or add a generation markup for that deployment. Confirm Managed MCP terms during onboarding.

• A Google key routes image requests to Google Gemini. The current Iris source uses gemini-nano-banana-2.1 by default and also supports gemini-3-pro-image and gemini-2.5-flash-image. Check server_status for your deployment; resolution and aspect-ratio choices depend on the model.
• An OpenAI key routes image requests to OpenAI image models, with size, quality, and background controls.
• A Google key routes video requests to gemini-omni-1.1-flash. Veo model names and Fast/Standard/Lite choices are no longer video controls.

Omitted resolution and aspect ratio inherit the source for edits and extensions. Google describes 1080p and 4K as upscaled outputs. Review actual media properties and product fidelity rather than treating a requested resolution as proof of quality.

Use a prompt alone, image with a prompt, or image plus last_frame. Describe whether the image is a product reference or starting frame, and describe the transition when supplying two frames. These roles are prompt instructions, not guarantees. Multi-shot prompts describe changes within a clip; they do not join existing videos.

The old reference_images, negative_prompt, enhance_prompt, person_generation, and seed controls are not exposed. Put exclusions, shot direction, and sound direction in the prompt. Iris does not expose uploaded-video sources, audio references, or voice editing.

Preflight, charges, and recovery

validate_video_plan checks the intended request before a billed submission. Review the prompt, settings, and available estimate before approving generation. A 720p estimate uses Google’s published video-output pricing; input and other billed tokens are additional. Estimates for 360p, 1080p, 4K, edits, and extensions are unavailable. An unavailable estimate does not mean free, and an extension estimate must not be calculated from only the added seconds.

Retained actual usage is separate from the estimate. Check returned usage after creation; do not promise a fixed price. After a timeout or pending result, reuse the same recovery_id with get_video_status. If submission is unknown, do not issue a fresh generation under a new ID: the provider may already have accepted the billed request.

MCP tools

The server exposes a small, clean tool surface for images and video.

HTTP deployments can return signed download URLs for generated assets. Video results also include an Iris video handle and a Range-capable URL. Confirm the URL is usable and remains valid long enough for the next step; images can move through Amazon's pull-based listing workflow, while video shots move to external finishing.

Chained edits and brand-mark fidelity

To iterate without re-uploading the source image, pass the prior interaction back in:

• Gemini: reuse interaction_id from a previous generate_image result (valid ~55 days on paid accounts, 1 day on free).
• OpenAI: reuse previous_response_id from a previous generate_openai_image result.

Brand-mark fidelity is a first-class review requirement. Instruct the model to preserve the original labels, logos and printed text, then compare the output against the source: generation does not guarantee an exact reproduction. A re-lettered or garbled label needs correction.

Image delivery to a seller listing

For supported Selling Partner listing-image updates, the connected client submits an approved image URL for Amazon to fetch. An HTTP Iris deployment can return a signed URL for a generated or uploaded image, so Amazon can retrieve the asset without a separate hosting step when that URL is publicly reachable and valid.

A signed URL expires, and the stored file can have its own retention policy. Check the returned URL and expiry, confirm that the deployment exposes a usable public origin, and allow enough time for Amazon to fetch the image. Confirm ingestion after the approved listing update; a generated image or a submitted URL does not prove the listing was updated.

Video delivery and retention

A completed video request returns an Iris handle such as agent-iris://videos/vid_N and a signed, Range-capable download URL. Google Files URLs require authentication; they are not permanent public sharing links. Download the deliverable through Iris and retain it according to your own asset policy.

Edits and extensions require an owned Iris-generated source, its originating credential, and valid retained interaction lineage. File availability and lineage lifetime are separate. The local 55-day lineage deadline is an upper bound; a shorter Google retention policy or unavailable upstream interaction can prevent continuation sooner.

Iris checks measured source duration before extension and returns the accumulated video. Inspect the actual result, including the transition and runtime. A requested duration or resolution does not establish the returned media's exact properties.

For longer deliverables, hand the editor the clips, in/out points, shot order, transitions, audio plan, and intended runtime. External editing handles assembly, final export, and required metadata. Confirm the finished file meets the chosen Amazon placement's current specifications, rights, disclosures, and seller eligibility before upload. Iris does not guarantee Amazon approval.

Limits and honest framing

• Amazon makes the final call. The skill materially de-risks and accelerates compliance, but it does not guarantee acceptance — Amazon does.
• Exact-white background. Earlier model tests observed near-white pixels; that observation does not establish the behavior of every current model. Prompting alone does not guarantee exact white, and changing file formats does not normalize the background. Where required, normalize it downstream and sample the corners and open background areas.
• Provider credentials. Image tools use the configured provider. Google video operations require a Google credential or an operator-configured Google fallback even when using OpenAI for images.
• Synthetic images must be honest. AI-generated product images are acceptable only when they realistically represent the actual product and never hide defects, change identity, color, scale, or contents, or add prohibited text or claims.
• Generated video must preserve the real product. Use a real product image to condition the clip and check the result against the seller-supplied references. A changed label, color, material, control, included part, or proportion fails the fidelity review and should be regenerated.
• People and regional restrictions apply. Check current provider restrictions when a prompt or image includes people. Iris has no selectable people-generation control.
• Iris does not finish or upload the listing video. It generates clips and edits or extends owned Iris videos. Joining clips, final export, applicable metadata, and Seller Central upload remain outside the server.

Related

• Google Gemini Omni model documentation — provider capabilities; Iris exposes the controls listed above.

• Connect to Managed Iris — endpoint, authentication, provider credentials and the first task.

• Amazon Agent Iris feature overview — the product surface at a higher level.

• Amazon Selling Partner MCP — publish generated images straight onto listings and verify ingestion.

• MCP client configuration — connect Claude, ChatGPT, Cursor, and other clients.