Creator Canvas

Creator Canvas is a single visual workspace for AI images, videos, and podcasts. Drag nodes, connect them to a Result, and Theo runs the full pipeline behind the scenes. Image and Video share a brand context layer (Style Guide, Logo, Brand Kit, Rules); Podcast is sound-first and uses its own podcast-native node set.

Three modes, one canvas

Image

Compose AI images on a node graph. Connect prompts, style guides, logos, brand assets, and rules to a Result node and generate.

Video

Storyboard a multi-scene AI video, attach a voiceover style and brand context, then dispatch the same render pipeline that powers Theo Video Mode.

Podcast

Compose an AI podcast episode from a topic (or full script), two hosts, and optional background music — all on the same node graph. Each podcast node has its own accent color so Topic / Script / Research / Host / Music / Audio / Episode are visually unambiguous.

Image mode

Compose AI images on a node graph. Connect prompts, style guides, logos, brand assets, and rules to a Result node and generate.

  • Drag in Prompt, Style Guide, Logo, Assets, Placement, Rules and Result nodes.
  • Pick a Theo image engine (Photo, Creative, Studio…) per Prompt node.
  • Use the Flow Builder to describe your output in plain language and let Theo wire the graph.
  • Mark nodes with priority on each connection so Theo knows which input dominates the final image.

Video mode

Storyboard a multi-scene AI video, attach a voiceover style and brand context, then dispatch the same render pipeline that powers Theo Video Mode.

  • Each Scene node is a single shot with prompt, optional reference image, and aspect ratio.
  • Voiceover nodes let you set narration text, the Theo Voice preset, and tone.
  • Music nodes generate a background score with Theo Score — pick a curated preset (Ambient, Cinematic, Upbeat, Lo-fi, Dramatic, Playful, Minimal) or write a custom prompt with bpm / density / brightness controls. Preview inline, then connect upstream of the Result node and the rendered track replaces the auto-cued score in the final cut.
  • Brand and Rules nodes carry over from Image mode so the look stays consistent.
  • The Result node calls the existing video pipeline and tracks Inngest job progress inline.

Podcast mode

Compose an AI podcast episode from a topic (or full script), two hosts, and optional background music — all on the same node graph. Each podcast node has its own accent color so Topic / Script / Research / Host / Music / Audio / Episode are visually unambiguous.

  • Topic node holds the title, focus, and a rich source-content blob the hosts will discuss.
  • Script node lets you write or have Theo build a complete multi-speaker script the hosts read verbatim — connecting it to a Research node auto-drafts a pre-script the moment you open the editor, and editing is gated on a Host being wired in.
  • Research node runs Theo Deep Research on a query and feeds the synthesized report into the Topic, the Script editor, or directly into the Episode.
  • Host nodes set persona and the Theo Voice preset (defaults to the standard Alex/Sam pair). Connecting a Host unlocks editing inside the Script editor.
  • Music node generates a background score with Theo Score — pick a curated preset (Ambient, Cinematic, Upbeat, Lo-fi, Dramatic, Playful, Minimal) or write a custom prompt with bpm / density / brightness controls. Preview inline, then connect upstream of the Episode node and the rendered track replaces the auto-cued score under the hosts. The intro / under-bed / outro toggles still control where the music plays in the final mix.
  • Episode node dispatches the existing podcast pipeline, polls for status, then unlocks Listen Now.
  • Generated graphs are auto-laid-out (topological column packer) so nodes never stack on top of each other.
  • Image-canvas brand nodes (Style Guide, Logo, Brand Kit, Assets, Rules) don’t apply to podcast mode — brand context is a visual concern that lives on the Image and Video canvases.

Shared building blocks

Style Guide, Logo, Assets, and Rules belong to the visual canvases (Image and Video). Research, Result, and Flow Builder are available across all three modes. Podcast mode adds its own per-node nodes (Topic, Script, Notes, Host, Music, Audio, Episode) and is documented in the Podcast section above.

Style Guide

Tone, mood, colors (Image / Video)

Logo

Brand mark / signature (Image / Video)

Assets

Patterns, swatches, fonts (Image / Video)

Rules

Hard constraints (Image / Video)

Research

Theo Deep Research inside the canvas (Podcast)

Result

Dispatches the AI run

Flow Builder

AI-built node graph from a prompt

Start from a template

Don't want to start from a blank graph? Open Templates → Creator Canvas for curated starting points across all three modes:

  • Image — Instagram posts, YouTube thumbnails, logo concepts, hero illustrations, product mockups, and more.
  • Video — product commercials, SaaS explainers, cinematic trailers, TikTok / Reels hooks, how-to tutorials, and presenter videos.
  • Podcast — 1:1 founder interviews, two-host deep dives, daily news briefings, comedy roasts, and narrative storytelling episodes.

Click Use Template on any card and Theo opens the Creator Canvas in the matching mode with the prompt pre-filled — review or edit, then hit Result to render.

Tips

Brand once, generate twiceDrop a Logo, Brand Kit, and Rules node onto Image mode and the downstream Video graph can reference the same brand context. The same nodes work in Image and Video. Podcast mode is sound-first and intentionally has no brand-context layer — voices, music, and script style live on the Host / Music / Script nodes instead.
Use the Flow Builder for the first draftType a brief like "A 60-second product launch video for our brand" and Flow Builder will scaffold the whole graph for you. Tweak the nodes after to taste — then hit Result to render.
How auto-save and the Canvas Hub workEvery Creator Canvas session auto-saves to its underlying chat conversation. The moment you open a canvas, a card appears in the Canvas Hub so you can find it again — even before you add a node. Titles update automatically from your first prompt or scene, and thumbnails come from the most recent generated image, video render, or published podcast cover. (One-time setup: run scripts/setup-creator-canvas-fields.sh so the hub flags + per-mode columns exist on Appwrite.)

Folders & multi-select in the hub

The Image, Video, and Podcast tabs of the Canvas Hub now support per-mode folders, multi-select, drag-and-drop, and bulk actions so a long list of canvases stays organized.

Create folders

Click + New folder on any per-mode tab. Pick a name, color, icon, and sticker — the dialog previews the folder face as you tweak it.

Drag canvases in

Grab any canvas card and drop it on a folder. The folder card glows in its color while you hover, and flashes green when the move lands.

Multi-select

Click Select in the header (or Cmd/Ctrl-click a card) to enter selection mode. The bottom toolbar lets you bulk move-to-folder or move-to-trash with one tap.

Bulk restore + delete

In the Trash tab, multi-select gives you a bulk Restore button and a guarded Delete forever action.

One-time setupRun scripts/setup-canvas-folders-collection.sh once to provision the canvas_folders collection and add the canvasFolderId attribute to chat_conversations. The hub stays usable on legacy schemas — folders just won't appear until the script runs.

FAQ

Will my existing Image Canvas projects keep working?

Yes. Image mode is the default Creator Canvas tab and runs the exact same node graph as before. All saved canvases, brand assets, and conversation history are preserved.

Can I switch modes in the middle of a conversation?

Yes. The 3-tab pill on the input bar (Image / Video / Podcast) flips the active node graph for the same conversation. Each mode keeps its own graph; the others stay safely persisted while you work.

Does brand context carry across modes?

Brand context applies to Image and Video only. Drop a Logo, Brand Kit, or Rules node on either canvas and the same context can be referenced in the other. Podcast mode intentionally has no brand context — a podcast is sound, not visuals — so the toolbar and Flow Builder won’t spawn those nodes there.

What about credits — does Video / Podcast mode burn through them faster?

Video and Podcast both call the existing background pipelines, so credit usage matches the amount you'd see from `create_video` / `create_podcast` in chat. Image generation credits are unchanged.

Related Articles

Was this article helpful?