Framework by NewForm vs Pencil

Framework by NewForm is web software for growth teams managing advertising creative. It launches creative tests, checks performance hourly, and evaluates results against KPIs to a stated 95% confidence threshold. When data is clear, it can pause underperforming creatives and scale winners. AI vision and language analysis examines ad elements and explains performance. Framework can use brand guidelines, compliance rules, performance data, competitor research, and a creative archive as context. It can turn account and competitor research into a ready-to-shoot brief or a Veo 3 prompt containing brand voice, disclaimers, and production notes. The product launches creative across Meta, TikTok, and YouTube, with Google included in cross-platform examples. It matches the same creative across platforms by visual content rather than filenames and combines performance data. NewForm says it targets advertisers running at least 20 creatives monthly and spending $50K or more per month.

Pencil is a web-based creative marketing platform that uses AI to generate advertising copy, images, and video ads. Its workflow runs from briefing and generation through editing, approvals, publishing, and evaluation. The browser editor handles text, image, video, and HTML, with exports in GIF, MP4, HTML5, and static formats. Pencil says a single template can produce thousands of variations and be adapted across markets through localization or translation. Listed generally available ad integrations include Meta, TikTok Ads, LinkedIn Ads, Google Ads, YouTube, Display & Video 360, and Campaign Manager 360. Users can create six ads for free through a guided experience, but full platform access requires a paid plan. Core costs $11 a month billed annually and includes 50 generations per month and one workspace. Growth costs $44 a month billed annually, with 250 monthly generations and unlimited workspaces. Pro has custom pricing and unlimited generations. Pencil says its security controls are SOC 2 Type II audited and that customer data is not used to train its models.