Updates

Each SKU's photos are a product lock: the reference set supplies only layout, model, scene and how details are presented, and close-ups are re-rendered from the locked product.

The product lock now calls out buttons on their own: the same number, colour and shape as your product photos, in close-ups too.

Batch Video Remake and Batch Image Swap are no longer enabled account by account — both are visible to every account by default.

With a person-less fabric reference, results no longer come back as model shots just because the SKU photos are worn by a model.

Fabric-detail frames now keep the reference's weave and presentation and change only the colour and pattern.

A seventh model joins the swap picker, with its frame snapped to the closest fixed ratio for each reference shot.

Reference colours no longer leak into swapped fabric shots, close-ups stop gaining buttons, and the limit on fine checks is now documented.

Eight inconsistencies in the swap prompt are resolved, and colour lock no longer pulls a new SKU back toward the old product.

The Materials batch bar gains Download selected: one asset downloads its original file, several are packed into a single zip of originals.

Image-to-image and smart-edit results gain Color lock: the overall colour cast is corrected against the reference and untouched areas are restored from it, free and in seconds.

Upload a finished set of shots plus the product photos of your other SKUs, press once, and every SKU gets the same set with only the product swapped.

A line's daily call cap now defaults to 10,000 (range 10–20,000); lines still on the old 1,000 default are raised automatically.

The creator pool filter bar gains an outreach filter: not contacted, contacted (any status), or an exact status.

Product Listing, Image Replica, Garment Pattern Variations and Pattern Transfer result cards gain the image workspace's Save to library button: a free independent copy that survives task deletion.

When you have folders, Save to library first asks which folder to file the copy under (defaulting to your last choice); without folders it saves directly as before.

Every keyword seed is now searched daily under four sort orders until the 100-product pool is full; the category rank only ranks and back-fills; own-shop products no longer enter the pool, and all pooled products get their creators looked up.

The product pool is paginated over every product and opens shop profiles; the sync adds product-detail and shop-profile steps; alerts name blocks by type and explain why the daily rank is not out yet.

Text to Image and Image to Image gain an AI prompt writer: describe the result in one line and the AI turns it, together with your reference images, into a professional prompt written the way the selected model's official guide recommends.

With the interface in Chinese, the AI prompt writer now shows an editable Chinese rendering of the prompt while the English version is what gets sent to the image model; the English interface keeps showing English.

The video workspace's preview panel gains a free Save to library button: the finished video is copied into your asset library so it can be reused as a reference video elsewhere.

A smart edit no longer overwrites the image: the original and every edit stay in a version history under the preview, and Set as current switches back for free.

A new ops agent: per category line (Vecardi, Viakeo) the board pulls TikTok Shop data every day and finds the creators actually selling the category, graded and scored on the golden metrics, with due-diligence and outreach records.

EchoTik's avatar links are private and cannot be loaded by the browser; the board now re-hosts every creator avatar on our own CDN and never shows a broken image.

The default daily call cap is now 1,000 and every creator of the day gets profile and video data, so 'latest video' is no longer limited to the top 20.

Refresh no longer answers 'today's data is complete' while profiles or videos are still missing under the current settings; it re-runs the sync for just the missing blocks.

The creator pool is now paged (20 per page, switchable to 50 / 100) instead of one long list.

Fixed column widths for the creator pool: numeric columns get their natural width, the creator and top-product columns share the rest, and nothing wraps.

The creator detail drawer is now a wide two-column panel so category performance, due diligence, videos and the outreach record fit on one screen.

In the creator drawer the contact e-mail now sits right under the account name instead of inside the due-diligence block.

Google Nano Banana 2 (Gemini 3.1 Flash Image) joins every image entry point: 8 / 12 / 18 credits at 1K / 2K / 4K, up to 14 reference images, new ultra-wide / ultra-tall 4:1 / 1:4 / 8:1 / 1:8 ratios.

Credit prices for the image models, Veo 3.1 and Gemini Omni now follow what Kie actually bills: GPT Image 2 halves to 3 / 5 / 8, GPT Image 2.5 1K drops to 3, Nano Banana Pro is 8 / 8 / 14, Nano Banana 2 is 5 / 8 / 12; all three Veo lines and every Omni tier come down with them.

The audio section of batch video remake now offers Voice-over: the AI writes the script from the reference video's pacing and your product info, each SKU only swaps the product wording; VO language and subtitles are selectable.

Product-info fields you leave empty are now read from each SKU's own images (what you type still wins for every SKU), and the voice-over is rewritten per SKU with the master's timing.

A new top-level "Batch Production" group beside AI Video, AI Image and AI Agents, opening with Batch Video Remake: one viral reference video + character photos + a scene photo + the product images of a batch of same-style SKUs, one click, and every SKU gets its own matching video with no interaction after launch.

Targets two failure modes, "the scene photo was ignored" and "a different person after the turn": actions broken down from the reference video no longer carry environment words, the environment comes only from the scene photo and scene text, and every SKU prompt gets a top-priority identity-lock block that re-anchors the same person on every turning shot.

The Reference to Video prompt wizard now shares the identity lock with batch video remake: a Seedance draft with a subject reference assigned gets a mechanical "Identity lock · top priority" block after linting, restated on every turning shot.

"Batch Video Generation" is now "Batch Video Remake"; each SKU group accepts up to 5 product images instead of 3.

The home page after sign-in drops the tool cards and shows your own balance, output, running tasks, spend and failure reasons — dark and light themes, phone-friendly.

The studio home's activity heatmap grows from 30 days to a full year in GitHub-contribution style: week columns, month labels, weekday labels and the year's task total.

The 21-image cap per set is gone: images are analysed 10 at a time, each batch is saved and shown as soon as it finishes, the set-level conclusions are compiled at the end, and a failed set can resume from where it stopped.

The sign-up bonus is back: a new account receives 50 credits the first time it opens billing or submits a task, shown as "Signup bonus" in the ledger; only the first account per sign-up IP is eligible.

Per-image breakdowns now cover first impression, style reference, presentation, composition, lighting, background & palette, main purpose, typography and copy.

Confirm Listing vs A+ before uploading; folders, zip files and images all take several sets at once, and picking a parent directory groups by sub-folder.

When a draft fails the health check, both the automatic revision and Fix all now change only the named shots and sentences and keep everything else (including your edits) verbatim; no charge when nothing can be improved.

When a patch targets findings such as a product shot without colour or material, it now carries your confirmed product facts and the product reference images so no colour is invented; other findings stay text-only and videos are never attached.

Viral video insights and listing image insights now require a tag on upload (e.g. "Top 100 apparel", "trending now"); lists and analytics filter by tag, older records can be tagged in bulk, and a single record can be edited on its detail page.

OpenAI's GPT Image 2.5 (released September 8) is live: Flare and Sunburst lines for both text-to-image and image-to-image at GPT Image 2 pricing, with native transparent output.

After you upload a video, enter its link and press Start, the finished breakdown is kept in the library and counted in analytics automatically — no manual confirmation.

The second ops agent: batch-upload Amazon listing and A+ images by folder, break every image down by composition, lighting, subject, typography, copy and color, keep the result automatically, and read the patterns on a new Image analytics page.

Listing image insights now takes the fixed export format directly: one folder per ASIN holding a listing-images zip and an A+ zip, unpacked automatically with the ASIN and an Amazon link attached.

The ops dashboard gains an Admin accounts card: grant or revoke admin by email without touching the database.

The first ops agent: upload a benchmark viral video with its link, break it down into 31 dimensions across 7 modules, review each one and keep the confirmed version.

A codebook turns every confirmed breakdown into 12 enumerated fields, and a new Analytics page shows the patterns across the whole library with filters, a library baseline and CSV export.

The viral insights library and analytics are now scoped per team: a master account and its members share one library, and accounts without a team see only their own records.

Per-user module exceptions in the ops dashboard now cascade to team members: a module enabled for the master account is available to its members without per-member setup.

Admins can add an account to a master account's team directly by email from the ops dashboard, or remove a member, without going through an invitation link.

New accounts no longer receive free credits; credits come from top-ups.

Team accounts gain a shared-pool mode: members spend directly from the master's credits and see the team's remaining balance, while the master reviews each member's consumption; one mode per team.

An owner account invites members by email, allocates and reclaims credits, and reviews each member's credit usage while creations stay fully isolated.

Balances and ledger rows no longer show floating-point residue such as 7.1e-15 or 1.2000000000000028: the ledger rounds to two decimals and historical data was backfilled.

Sign up and sign in with an email code — no password needed — plus a confirm-password field and automatic hand-off to the studio once a password account is verified.

Product Listing Set adds an Amazon A+ Content platform — six categories planned against the official module specs, minimal-text images with paste-ready module copy, per-module aspect ratios and module-size downloads.

MiniMax H3 prices are halved — 768P 16→8 and 2K 26→13 credits/second, per-image surcharge 8→4; synced the same day by the daily price check.

Garment Pattern Variations now focuses on one product per task, and adds three optional strict references — flat-lay base, model reference and scene reference: provided images are followed exactly, otherwise AI matches the product.

Reference to Video adds a Reference image tips gallery — green checks for good references, red crosses for ambiguous ones that make the model guess and mismatch your product.

Scene, action, camera and audio sections each gained a one-line Section notes input: state your constraints before generating and the candidate plans are designed around them from the first draft.

A new free-form requirements box below the Audio section: write anything extra and the draft fulfils every item; brand text and timelines written there count the same as in any other section.

A batch of official-guide alignments: Seedance 2.5 multi-angle portraits enabled, hand-off ownership and spatial-anchor disciplines, official sound notation, whole-second timelines, an edit-routing disambiguation note, and asset-count hints.

Every requirement in the unified brief now lands in every card's prompt: numbers verbatim, style names and structural selling words in explicit English, with a deterministic number audit and one automatic repair pass.

A new Logo-reference image role plus letter-by-letter brand-text disciplines and health checks; verified with real renders: a lettering image on Seedance 2.5 / 2.0 Standard reproduces invented brand words letter-perfectly.

Static product-display shots in a reference video are flagged in the shot picker and, when kept, anchored to your product image as the first frame instead of being mirrored; the wizard reminds you to assign a product reference.

The wizard's free media analysis now detects the product category (Apparel / General) and switches the selector for you; a manual pick is never overridden.

Turn a same-category product demo into a video of your own product: shot detection judges the source, the picker selects one span, and the wizard writes a video-edit instruction that swaps objects one by one while keeping people, motion and camera.

Product-display shots in a reference video are now re-dressed from your product images instead of forcing a product image as the opening frame; the health check blocks unrequested first-frame anchoring.

Pick several reference shots from a long video; view and retry keep the shots; the prompt wizard fills goal, scene, action and camera in one click.

Every wizard draft is checked item by item (asset references, product anchors, subject consistency, timeline, voice-over length, render parameters); findings can be fixed by hand or regenerated in one click, and the server auto-revises once before delivery.

Rate every finished video 👍/👎 with issue tags; feedback links to the wizard recipe behind it and feeds an ops dashboard that shows where prompt quality breaks.

A public pricing page with four one-time credit top-up tiers, bonus credits on larger packs, and a contact-to-top-up flow via corporate bank transfer.

Drop files from your desktop straight onto any material area — every module, one gesture.

A new agent that modifies, adds, removes, re-voices, extends and splices existing videos on the Seedance family.

Seedance 2.0 Fast / Mini and MiniMax H3 prices drop, with quotes now matching what is actually charged; a daily price check keeps future changes synced same-day.

The wizard can now write per-shot voice-over scripts from your product info, or design a matching background-music plan; with a reference video it can also follow the source audio.

Scene-first cascade, explicit failure notices, guaranteed scene info, retry-as-new-session and full detail-image utilization — a day-one systematic polish of the wizard.

Seedance renders gain a third generation route at identical prices, and route selection moves off the form entirely — the platform now schedules the route automatically, switchable by operations within a minute and frozen per task.

A five-step interactive wizard in the Video workspace turns your intent into a structured prompt that follows the official Seedance prompt guides — with free AI media analysis, per-dialect writing rules, and a 2-credit generation fee.

The Video workspace gains Alibaba's HappyHorse 1.1 (nine aspect ratios, all three modes); every creation page now carries the same “New task” button; Motion Transfer defaults back to Kling; Seedance defaults back to the primary route; and three buttons that phone-width cards used to clip are fixed.

Motion Transfer adds Wan 2.2 Animate as its default model — non-human subjects, source videos up to 120 seconds, roughly half the price, and the video's own scene without preprocessing. Text to Speech also reworks pause markers and emotion tags per tier.

Upload one character photo and one motion video to make the person in the photo perform that motion — optionally in the video's own setting. Kling 3.0 and 3.0 Turbo join the video workspace, and Text to Speech now remembers your last voice.

The audio agent lineup opens with Text to Speech — turn scripts into natural spoken audio with curated voices, pause markers, emotion tags and AI script polish, fully wired into My Creations and billing.

Clone your own voice from 1–3 reference recordings and use it right inside Text to Speech — strictly private to your account, with instant previews, slot management and automatic refunds.

Garment Pattern Variations and Pattern Transfer can now generate an independent, print-ready pattern vector file (4K, 300dpi, transparent PNG) from any successfully rendered flat-lay.

Garment Pattern Variations and Pattern Transfer let you choose which cards continue into model-shot and pattern-vector batch generation, instead of always rendering every ready card.

Seedance 2.5 now offers the same per-task route choice as the rest of the Seedance family, verified end to end on the alternate route.

Admins can now control which business modules each user sees — restrict individual accounts or open a new module to selected users only, configured from the operations dashboard.

Seedance 2.5 joins Video, Storyboard and Viral Video Remake with verified 480p/720p rendering, dual-tier pricing and a fixed route.

The AI Agents sidebar and dashboard home are now organized into a collapsible two-level nav by media type, with a reserved place for upcoming audio agents.

A new 6th agent migrates one pattern pixel-faithfully onto every colour of the same garment style, with free extraction, colour-mode control and the same two-tier flat-lay/model-shot billing as Garment Pattern Variations.

The product name now makes its garment-pattern workflow explicit across the app and user manual.

Open browser tabs now tell users when a new AI Studio release is ready to refresh.

MiniMax H3 joins Video, Storyboard and Viral Video Remake with verified multimodal controls, input-aware pricing and documented continuity limits.

Edit flat-lay and model-scene prompts in one place, with only the required images regenerated.

Organize one shared Materials library with folders, recent use, favorites and Unfiled—without copying or silently moving files.

Free source analysis is now a durable task, with paid generation confirmed only after review.

New Garment Pattern Variations tasks now start model shots at a 1:1 ratio.

GPT Image 2 HD is now the default alongside Nano Banana Pro and Seedream 5 Pro.

Sign-in now opens directly on email and password, while Google and Settings phone binding remain available.

A narrower desktop sidebar and clearer Garment Pattern Variations model names make every shared workspace easier to use.

Each new variant now receives a distinct fictional English brand name and matching typography.

Seedance now offers a per-task route choice, while the alternate route safely imports every reference into its material library before generation.

Every generation workspace now exposes Adjust & retry immediately after submission while the original task keeps running.

Eligible test accounts can compare the stable original-video path with depth-only motion transfer and submit per-segment quality feedback.

Review Depth render requirements in Chinese or English while generation keeps its model-optimised English prompt.

Every paid image, video and agent action now gives the same clear low-balance message without silently disabling the action or starting a task.

Upload one garment and a few pattern references, and AI fissions 10–20 variants — flat-lay shots for review first, then model shots for the ones you approve.

Upload any mix of reference photos and AI free-analyzes each one's role and style — one submit now drafts the plan and renders every flat-lay automatically.

Product Listing Set, Image Replica, Viral Video Remake and Video Upscale ship, alongside an English/Chinese user manual, remembered My Creations filters, and a unified content font.

Video Upscale gains an Aliyun model — a fixed 2× upscale with quality enhancement, selectable alongside Volcengine and priced by source resolution.

A full e-commerce storyboard-to-video pipeline, faster galleries, an image zoom viewer, and bulletproof refunds.

AI Studio is live — video and image generation, credits, an asset library, and My Creations.