Skip to main content

The media studio

Images and video live in Images & Videos (sidebar), one library the whole workspace shares: generated, uploaded, or dragged in. The editor pulls from it (long-form pieces insert images at the cursor with the toolbar's Image button; social pieces attach media through the Media bar), the campaign canvas generates on it with media prompt cards, and images and video can also publish straight from the library.

The create wizard​

+ Create walks three steps: Type (image, video, or template import), Style, Prompt.

  • 87 named styles across photo, illustration, vector, and emblem families: Hard Flash, Neon Calm, Linocut, Pop Art, Pixel Art, and 82 more. Each style is a tuned replication prompt, so the look holds across models instead of living inside one provider's preset. Skip the step for a default on-brand look.
  • Your brand style leads the wall: a visual identity derived from your site during onboarding, ready before your first prompt.
  • Prompts are grounded automatically: the piece's opening, your brand name, palette, positioning, and top audience ride along, so images come out consistent without prompt-writing.

Generation is asynchronous and survives the tab closing; the library shows progress until the render lands. Generated media is stored by Reactor, so URLs do not expire.

Image models​

Two, switchable per generation:

  • Recraft: the default for styled art.
  • Nano Banana: photoreal work and true edits. Fresh generations take your brand mark as a reference image, and iterating edits the actual pixels ("lighter background, remove the text") while preserving everything else.

Pick the image type (blog header, social, in-article, graph) and size preset; an image linked to a piece as its blog header becomes the post's hero automatically on publish.

Video​

Video is Runway, image-to-video: pick a seed frame from your library (or let the wizard generate one), describe the motion, and get 5 or 10 seconds back in landscape, portrait, square, or ultrawide. Any image's detail page also has Convert to video with the same controls. Video types: social clip, promo, explainer, teaser.

There are no video templates, scripts, or voiceover; clips are visual.

Templates​

Import Figma frames as image templates: layers named #headline, #image, #logo, #subhead, or #cta become editable slots, untagged text layers are inferred automatically, and everything else bakes into a fixed background. Renders pull your real palette, brand fonts, and logo, and the AI writes the slot copy from one brief, sized to fit each box. Re-sync pulls a changed frame from Figma and your bindings survive.

Connect Figma under Integrations, then pick a template explicitly when generating or leave Auto template on to use the default for that image type and platform.

Uploads​

Upload (or drag files anywhere onto the library) brings your own media in: images and video up to 512 MB. Video uploads capture a poster frame automatically, and anything uploaded from inside the editor also lands in the library for reuse.

Publish from the library​

Any image can post directly to X and LinkedIn, and any video to X, LinkedIn, and YouTube, with an optional caption (an empty caption posts just the media). Each channel can be published to once per asset from the library; for reposts and threads, attach the media to a piece in the editor.

When media rides a piece instead:

  • YouTube uploads the piece's video (it lands private; flip visibility in YouTube Studio)
  • X and LinkedIn attach the piece's images (up to 4 on X, 9 on LinkedIn) or one video, never both
  • Instagram publishes the pinned image with the caption (attach the one to post in the editor's Media bar; without a pick, the newest linked image publishes)

Alt text​

Every generated image gets alt text automatically, derived from the image's subject (or a template's text content) and trimmed to 150 characters. It is stored with the image and attached where platforms support it, such as X media uploads.

Over MCP​

generate_image (with a style key, a template id, or "auto"), generate_video, list_media_styles, list_image_templates, render_template, and import_media cover the studio headlessly; poll get_image / get_video for async renders.

Cost​

Media draws from your AI allowance: about $0.06 per image and about $0.38 per 5 seconds of video. Media always runs on Reactor's provider keys (there is no BYOK path for it), and once the allowance is exhausted, media generation needs usage-based billing; there is no free fallback as there is for text.