Anvil logoAnvil
Product Updates & Releases

What’s new in Anvil

New features, engine improvements, model rollouts, and fixes shipped directly to the Anvil platform and CLI ecosystem.

Current Version

v1.32.0

Last Updated

October 27, 2026

Total Releases

65 Updates

System Status

Operational

3 weeks ago
v1.28.0MINOR
Platform Update
Added
  • anvil-studio now edits videos you recorded yourself. Say "edit my recording talk.mp4" and it transcribes the speech first, through Requesty's Whisper large-v3 (about $0.03 an hour of audio) or on your machine if you say "keep it local". It detects the spoken language unless you name it and fixes misheard product names such as "Cloud Code" for Claude Code. Then it plans the edit with you: captions only, cards on your own cut, or the full edit, which removes dead air, restarts and stumbles and adds captions, term and code cards, chapter titles, b-roll and a few animation moments. Every cut is listed in the plan so you can put one back by naming it, and a real screen in the b-roll always comes from your own recording or screenshot, never a generated one. You get the video, an SRT in the spoken language (another language is translated from it with the same timing) and a chapter list ready to paste into YouTube.
  • anvil-studio now makes cinematic ad films from generated video clips. Ask for "a cinematic 20-second ad for the coffee subscription" and each beat becomes a Veo 3.1 Fast clip, about $0.30 for 8 seconds at 720p, made from that beat's approved still so the clips look like one shoot. Words, the logo and the ending stay as sharp layers on top, a person never appears in two clips, and every clip is checked for warped hands, drifting labels and other generated-video flaws before it is used. A film run has a $5 budget by default, every clip is priced before anything is charged, and a clip that fails twice falls back to its still with motion. Without a kie.ai key you still get the film, built from stills.
3 weeks ago
v1.27.1PATCH
Platform Update
Changed
  • ai-humanizer has been rewritten. It now looks for 26 specific signs of AI writing, strongest first, taken from Wikipedia's "Signs of AI writing": contrasts like "not just X, but Y", one-line closers, forced lists of three, dashes everywhere, inflated claims, sales language, stock AI words, bold labels and leftover chatbot phrases. It keeps every fact, name, number and date in your text and adds none of its own. Give it a sample of your writing and it matches your voice, dashes included. Paste text to get a draft, the patterns still left and a final version; name a file and it rewrites only the prose, leaving code, commands and links alone; ask it to check or review text and it lists what it would fix without changing anything.
4 weeks ago
v1.27.0MINOR
Platform Update
Added
  • anvil-studio now makes slide decks. Ask for "a pitch deck for investors, about 10 slides" or "slides for Friday's team meeting" and it plans a 16:9 deck of 8 to 20 slides around one ask, with a headline on every slide that says its point as a sentence and speaker notes for what you will say over it. You get a PDF and a PowerPoint file; the PowerPoint shows each slide exactly as designed, with your notes in the notes pane, but its text is not editable there, so changes come back through the studio. Before you see the deck it checks that the headlines alone tell the talk, that every slide reads from the back of the room, and that the last slide carries your ask instead of "Thank you". "Slides" for Instagram or LinkedIn still makes a swipe carousel, and slides for a YouTube episode still go to youtube-content.
  • anvil-studio now draws charts and diagrams from your real numbers instead of asking an image model to invent them. A slide, poster or carousel can carry one big number, a few sorted bars, a single trend line, a share of a whole, a before-and-after, or a simple process or structure, each drawn as sharp vector graphics in your brand colours with its unit and source underneath. If a number you need is missing, it asks you once; it never makes one up.
4 weeks ago
v1.26.0MINOR
Platform Update
Added
  • davinci can now help you write a design brief before anything is designed. Say /davinci brief landing page, add --brief, or ask in your own words ("I want a brief to revamp my landing page with motion and a scroll tour"). It fills in what it can from your code, project wiki and design system, asks you at most four questions with a recommended answer each (so "go" is enough), then offers two or three starred concepts, such as the scroll tour, before writing anything. The finished brief lands in prompter/briefs/ and you change it one section at a time. It works for any page, flow or component, not only landing pages, and builds no previews or code. A later davinci run that finds the brief designs from it without asking its discovery question again.