See the recipe,not just the result.
Every step of your workflow is a card you can see, rewire and rerun. Images, video, voice and words on one canvas.
ESPick an edit. Only the steps it touches will run again.
Describe it. Get the workflow.
Tell the copilot what you want and it adds the steps and wires them for you. Suggest mode waits for your yes on each change; Auto just builds.
⌘JMake a vertical ad from my product photo, with a Spanish voiceover and captions.
Added six steps and wired them. The brand kit feeds the scene and the script. Run it, or ask me to change anything.
- Upload imageYour product photo
- Brand KitYour saved brand
- Generate imageScene, 9:16
- Generate video5 seconds of motion
- Generate speechSpanish voiceover
- Add video subtitlesCaptions from the script
One prompt. Three models.
Wire one prompt into three generators and compare them side by side. Keep the one you like; the rest of the workflow doesn’t move.
“An unbranded faceted glass perfume bottle with amber liquid on a mirror-like salt flat at dusk, warm low sun from the left, long shadow, film grain.”
Nano Banana Pro20.1 credits
Seedream 4.56 credits
Flux 2 Pro2.3 creditsMade in Orisu in one run, October 2026. Swap the model on a node and the rest of the workflow stays wired.
Build it once. Reuse it forever.
A finished workflow is a template. Run it again with a new brief, a new product or another client’s brand kit.
Anyone with the link sees the workflow read-only: every node, every model, every output. The recipe travels with the result.
Start from a templateRunnable workflows for statics, UGC, product films and localization.You build it. They run it.
App Mode turns any workflow into a simple form for the team. The canvas stays yours to change; they get a tool that always makes on-brand work.
See App ModeRuns that loop, and wait for you.
One workflow can make ten variants, and it can stop halfway for a person to choose. Both live on the canvas as ordinary steps.
- Seven days, no break-in
- Lighter than it looks
- Built for long miles
Variant 1 of 3
Variant 2 of 3
Variant 3 of 3
✓
✓
✕
✓What’s on the shelf.
Every card you can drop on the canvas, grouped the way the studio colors its wires. Image ports are green, video purple, audio orange.
- Generate image
- Edit image
- Change camera angle
- Change aspect ratio
- Remove background
- Upscale image
- Vectorize image
- Generate video
- Generate lipsync
- Edit video
- Merge videos
- Add video subtitles
- Remove video background
- Upscale video
- Extract video frame
- Generate speech
- Change voice
- Dub media
- Generate music
- Generate sound effects
- Add sound effects
- Transcribe audio
- Generate text
- Prompt concatenator
- Split text
- Brand Kit
- Brand voice
- Brand guidelines
- Brand docs
- Brand context
- Upload image
- Upload video
- Upload audio
- Product from website
- Meta Ads scraper
- TikTok scraper
- Google Drive import
- Google Drive export
- Human review
- If / Else
- Router
- Switch
- List selector
- Composer
The details that matter on a busy canvas.
What keeps a canvas usable once it has thirty nodes and three people on it.
- Groups
Select nodes and convert them to a group, then name it and pick one of seven colors.
- Comments
Threaded notes pinned to a node or anywhere on the board. They stay private to your workspace.
- Versions
Save a version before a big change. If two people edit at once, Orisu asks which to keep and files the other.
- Rerun one node
Run a single step again without touching the rest. Edits made during a run apply to the next one.
- Cost on every node
Each generation node shows its estimated credits, and the run shows the total before you press it.
- Everything is saved
Every output lands in your asset library. Deleted assets wait in Trash for 30 days.
- Share and clone
A read-only link shows the whole canvas. A signed-in viewer can clone it; keys and credentials are stripped.
- Inside Claude and ChatGPT
Over MCP the canvas renders right in the chat with live status, and opens in the studio in one click.
Before you ask.
Do I need to know how to code?
No. You connect cards with wires. Start from a template, or ask the copilot to build a workflow for you and edit it from there.
Which models can I use?
More than 100 image, video, voice and text models, all behind the same ports. Swapping one never means rebuilding the workflow.
What happens when I rerun a workflow?
Only the steps you changed, and the steps after them, run again. Everything else is reused, so you only pay for what changed.
Can my team use a workflow without seeing the wiring?
Yes. App Mode turns any workflow into a simple form: inputs on top, a button, outputs below.
Can I share a workflow with a client?
Yes. A share link opens a read-only view of the whole canvas, outputs included.
Can the copilot build a workflow for me?
Yes. Press ⌘J and describe what you want. It adds and wires the steps, makes small edits on request, and can run the workflow, check the result and refine it. Deletions always ask first.
Can one step run for a whole list?
Yes. Give a step a list, such as hooks, products or languages, and everything after it runs once per item. Each variant is tracked on its own.
Can someone approve results before the run continues?
Yes. Add a Human review step. The run pauses, a person picks the items that continue, and only those move on.
Where do my outputs go?
Every output is saved to your asset library automatically. Deleted assets stay in Trash for 30 days.
Can I see the canvas inside Claude or ChatGPT?
Yes. Connect Orisu over MCP and the canvas renders read-only in the chat with live status, plus a link to open it in the studio.