Changelog What's new across the catalogue.
4 Nov 2026 edit_video now runs Sume video models — lip-sync to new audio, captions, dubbing, upscaling, depth maps, motion transfer and object swap. Pass a sume- model key from generate-video list_models (type video_to_video). Added Sume as an image provider — 25 models including GPT Image 2.5, Nano Banana Pro, Seedream 5.0, Ideogram 4.5, Recraft 4.1, Grok Imagine 2 and Higgsfield Soul. Use a sume- prefixed model key. Added Sume as a video provider — 48 models including Seedance 2.5, Kling 4, Veo 3.1, Wan 3.0, MiniMax H3, Vidu Q4, HeyGen Video 1 and talking-avatar lip-sync (OmniHuman, Kling Avatar, InfiniTalk). Use a sume- prefixed model key. 28 Oct 2026 Added similar_ads, find_lookalikes, winning_hooks, creative_benchmark and enrich_domains Look up brands by website or Facebook page, and pass a website instead of an advertiser id to search_ads, find_peers and compare_brands get_ad can now return on-screen text, story beats and a link to the creative 22 Oct 2026 Swap now accepts a written fabric description when you do not have a swatch Added swap_many: compare up to four fabrics on the same photo in one call Added recolor: change furniture colour by name, up to four colourways at once Added make_swatch: create a seamless swatch from a description Optional 2x upscale on every result, plus target and pattern scale controls 16 Oct 2026 Reading documents and classifying transactions now uses Gemini 3.8 Flash. Meeting notes now use Gemini 3.8 Flash. Claude is no longer queried; checks cover ChatGPT, Gemini, and Perplexity. Research synthesis now uses Gemini 3.8 Flash. Document analysis now uses Gemini 3.8 Flash. Drawing analysis now uses Gemini 3.8 Flash. Caption writing now uses Gemini 3.8 Flash. Multi-engine visibility no longer includes a Claude search engine; it covers Perplexity, OpenAI search, and Gemini search. Email screenshot analysis now uses Gemini 3.8 Flash. Resume writing now uses Gemini 3.8 Flash. Answer drafting now uses Gemini 3.8 Flash. 7 Oct 2026 Added optional LiteScrape lookup actions with source identifiers and pagination monitoring_stations works again and lists the nearest stations first; search radius is now up to 25 km Added public storefront search, detailed app pages and competitor reviews Added ebook and audiobook storefront browsing and charts Added optional LiteScrape lookup actions with source identifiers and pagination Use Firecrawl v2 JSON extraction per product page, preserving each source URL and provider-credit costs. Retry temporary exchange-rate provider outages and keep error messages bounded. Added optional LiteScrape lookup actions with source identifiers and pagination Added optional LiteScrape lookup actions with source identifiers and pagination EPA searches now apply facility-name filters, normalize program codes, honor result limits and return the matching total. Added movie and television storefront browsing and charts Added public maps, reviews, business posts and live busyness across more place directories Added public app and game browsing, product details and competitor reviews Use Firecrawl v2 JSON extraction instead of the deprecated extract endpoint. Concise responses now honor fullContent and maxPreviewChars, returning the requested page text without a second call Structured extraction now uses synchronous Firecrawl v2 JSON scraping, with ordered results per URL and provider-credit accounting. Concise responses now honor fullContent and maxPreviewChars, returning the requested page text without a second call Added an explicit rendered-page source with selectors, cookies and browser controls