Get started with Lumina.
A fast quick-start for Lumina's image and video generation, plus the Canvas workflow basics — including a downloadable 30s demo workflow. For the full continuously-updated guide, contact our team.
Best practices for image & video generation.
Lumina pairs Seedream image models with Seedance 2.0 video to cover the full creative pipeline. For first runs, start with Seedream 5.0 Lite for stills and Seedance 2.0 for video.
Product Image Set Generation
One product photo → a full set of marketing-ready images.
- 01Upload the raw product image + bullet selling points.
- 02Agent (Doubao 1.8) writes production prompts automatically.
- 03Generate hero shot, selling-point visuals, 4-up angles, lifestyle scene, detail close-ups.
Cinematic Product Ad — Sunset Style
One product photo → a 20s narrative ad.
- 01Doubao writes a 20s shot-by-shot storyboard.
- 02Seedance 2.0 generates the 0–10s clip.
- 03Extract the last frame, feed it as the first frame for the 10–20s clip.
- 04Merge clips + audio for a seamless cut.
Dual-Character Product Demo
Two consistent characters demonstrating your product.
- 01Doubao splits the script into per-character descriptions.
- 02Pre-generate each character as a reference image (locks identity).
- 03Generate two video segments using the locked references.
- 04Merge into one multi-shot demo.
Campaign Creative at Scale
Same idea, many formats, ratios and channels — built once.
- 01Build the workflow once (brand reference → composition → output).
- 02Use Batch Generate to fan out across formats and styles.
- 03Reuse the workflow as a template every campaign cycle.
30s Two-Character Short
Anime, 3D or live-action style — short narrative video.
- 01Input storyline + style → Doubao generates the script.
- 02Extract character descriptions, generate per-character references.
- 03Generate two video segments and merge into one 30s clip.
IP Storytelling — Pet Radio Host
Give a mascot or pet a consistent voice across episodes.
- 01Upload IP image + idea → Doubao writes a ~30s script.
- 02Lock a fixed voice ID for consistency.
- 03Split script via regex, generate two clips, stitch with last-frame trick.
Four habits that save hours.
- 01
Start light
Pick a smaller model when iterating, then scale up for the final render.
- 02
Always save
Add a Save node to every workflow so outputs aren't lost.
- 03
Run partials
Select a node and click ✨ instead of re-running the entire workflow.
- 04
Steal templates
Use templates from Inspiration first; build from scratch once you know the pattern.
Lumina Canvas — workflows, not one-offs.
Canvas is Lumina's flagship: an infinite visual workspace where text, images, video and audio nodes chain into a repeatable pipeline. Build once, run end-to-end, scale across inputs. Drop in an Agent node to control it in plain language.
Add a node
Click + in the toolbar, or double-click anywhere on the canvas.
Upload an asset
Double-click canvas → File, or use a Load Image / Load Audio node.
Connect nodes
Drag from an output port to the next node's input port.
Run full workflow
Click Generate at the bottom of the canvas.
Run a single node
Select the node, then click the ✨ icon above it.
Save workflow
Use Cloud Save after a successful run — reopen anytime.
Pet Podcast — 30s 2-clip workflow
- 01IP image + ideaInputs
- 02Doubao 1.8Script generation
- 03Seedance 2.0Clip 1 (0–15s)
- 04AIPGetVideoAnyFramesListLast frame → first frame
- 05Seedance 2.0Clip 2 (15–30s)
- 06Merge VideoStitched 30s output
A ready-to-import Canvas workflow that turns an IP image + a short idea into a 30-second podcast-style video. Doubao writes the script, Seedance 2.0 generates clip 1, the last frame becomes the first frame of clip 2, and both are stitched with a locked voice for consistency.
- Step 01
Download the .json file below.
- Step 02
Open Lumina → Canvas → blank canvas.
- Step 03
Drag the .json onto the canvas to clone the workflow.
- Step 04
Swap inputs (your IP image, your script idea) and click Generate.
Need the full, latest guide?
This page covers the essentials. Our team maintains a deeper, continuously-updated playbook — workflows, prompts, model picks and onboarding tailored to your use case. Drop us a note and we'll send it over.