Generator
AI Character Generator on Astorie
Generate the master portrait, then reuse it across every scene and shot. Drop the character description, lock the canonical portrait on Nano Banana 2 or Midjourney, fan out turnaround / expression / outfit views on Flux Kontext — the face stays on-model because every variant inherits the same anchor. Built for indie game studios prototyping the protagonist, novelists building a recurring-character bible, and AI influencer producers at week one of a twelve-week pipeline.
What you can generate
- Original game protagonist or NPC base portraits
- Recurring novelist character portraits for serial fiction
- Brand mascot characters that anchor the visual identity
- Webtoon and manga lead character bibles
- Tabletop campaign cast — players and NPCs anchored to a shared style
- AI influencer personas with locked face and shifting wardrobe
- Expression sheets covering five to six emotional states per character
- Outfit variations — default, alt, seasonal — from the master portrait
Best Astorie workflow
Why this is more than a one-shot generator on Astorie.
- Nano Banana 2 or Midjourney generates the master portrait once. Flux Kontext runs every outfit, expression, and pose variation off that anchor instead of re-rolling the face per shot.
- The character is reusable across image and video on the canvas — the same anchor that produces the storyboard frame also feeds the Vidu or Kling O3 video shot.
- This generator is the upstream of /features/ai-character-consistency and /workflows/multi-shot-short-film — every downstream pipeline reads the same portrait node.
- Refine the still to a hyper-clean state on Flux Kontext or Nano Banana 2 before feeding it into a video model — raw portraits drift more than refined ones.
- Frame the deliverable honestly — a concept-stage bible for the artist team to clean and finalize in Procreate or Photoshop before rigging in Spine, Live2D, or DragonBones, not engine-ready production sprites.
- Save the canvas as a character template — the next book, episode, or campaign reuses the entire chain by swapping only the costume and scene anchors.
Recommended models
midjourney
imageEditorial composition for the master portrait — strongest cinematic look development for hero character views.
nano-banana-2
imageStrongest face-locker — the canonical anchor model for recurring-character identity across scenes and shots.
flux-kontext
imageOutfit and expression edits from the master while preserving the anchored face.
flux
imageHigh-fidelity rendering for finished hero views and final-quality character cards.
gpt-image-2
imageCleanup pass for fine detail, color consistency, and edge polish before downstream video.
Prompt examples
Wiry teen runner, mid-teens, navy hoodie, copper-rimmed goggles pushed up on forehead, signature scar on left cheek, cyberpunk-noir lighting, neutral backdrop, 4:5 master portrait framing.
YA-fiction protagonist master portrait — strong identity markers (goggles, scar) anchor the recurring face.
Grandmother witch, 70s, deep purple cloak with silver embroidery, kind eyes with crow feet, soft daylight from window, weathered hands resting on a wooden staff, 4:5 framing.
Cozy-fantasy recurring character; the staff and cloak read as wardrobe constants across scene swaps.
Cyberpunk hacker, mid-20s, neon-magenta hair tied back, jacket layered with embroidered patches, rain-slick alley behind, dramatic backlight, 4:5 framing.
Cyberpunk hero — neon hair and patch jacket are the visual anchors that survive scene swaps.
Brand mascot fox, friendly stylized cartoon, cream-colored fur with white belly, oversized round eyes, plain neutral background, 1:1 framing for icon use.
Brand mascot baseline — illustrated style with locked color palette for cross-channel reuse.
Game protagonist warrior, late 20s, plate armor with cyan magical accents, scar across right brow, determined expression, three-quarter turn, neutral grey backdrop, 4:5 framing.
RPG hero base portrait — armor accents are the visual signature that survives video shot fan-out.
Expression sheet, same character as master, five expressions: happy, angry, surprised, focused, exhausted — same lighting, same wardrobe, identical framing per cell, 16:9 sheet.
Five-pose expression sheet generated off the locked anchor for storyboard and reaction-shot use.
Outfit variation, same character as master portrait, swapped to seasonal winter wardrobe — wool coat, knit scarf, leather gloves — same lighting, same backdrop, 4:5 framing.
Wardrobe swap from the anchor; identity stays locked while the outfit changes.
Three-quarter turnaround, same character as master, viewed from front-three-quarter, profile, and back-three-quarter — same wardrobe, same lighting, identical framing per cell, 16:9 turnaround sheet.
Turnaround sheet for downstream video shot blocking and 3D handoff.
Turn this output into a workflow
Generation is the first node — here's where to take it next.
Follow /how-to/generate-consistent-character for the per-model recipes behind the master-portrait anchor.
Wire the locked portrait into /features/ai-character-consistency to keep the face on-model across cross-modal output.
Chain into /workflows/character-consistency for the full image-to-image-to-video pipeline.
Use the master in /workflows/multi-shot-short-film as the recurring-protagonist anchor across scenes and cuts.
Pair with /prompts/image/consistent-character-prompts for the catalog of paste-ready identity recipes (turnaround, outfit, expression sheets).
Related features
Related how-to guides
Related prompts
Related workflows
Related reading
Frequently asked questions
Will my character hold identity across a full bible of views?
With a locked master-portrait anchor, yes — far better than re-typing the description into a dozen sessions. Even so, expect some drift on small details (a button on the jumpsuit, the goggle strap) and review every view before declaring the bible complete. For commercial games and serial novels, the character artist usually does a final cleanup pass on the AI output before the bible ships to the rig artist.
How do I keep the face on-model across scenes?
Use one canonical reference image. Never feed the model multiple "ideal" portraits — it averages them and the face drifts. Anchor on Nano Banana 2 and run every variant through Flux Kontext from that anchor. Give each named character its own anchor — the world grows by adding anchors, not by re-prompting.
Can I do style transfers of known IP characters?
For personal study and pure reference, image models will produce those styles, but for commercial release the answer is no. Style mimicry of Pixar, Ghibli, Marvel, Disney, or a living artist is trademark and right-of-publicity territory, and recent litigation has narrowed the safe zone considerably. Generate original characters and run them through Flux Kontext to land your own visual signature instead of a derivative one.
What about the video step?
The character lives across image and video on Astorie. The same anchor portrait that produced the storyboard frame feeds Vidu, Kling O3, or Kling Avatar for the video phase — see /workflows/character-consistency for the full pipeline.
How do expression sheets work?
Generate the master portrait first, lock it as a reference node, then run a five-cell sheet prompt that holds wardrobe and lighting constant while the expression changes. Flux Kontext is the cleanest model for this pass.
Can I use this for a tabletop campaign cast?
Yes — generate one master per character on Midjourney, lock each as a separate anchor node, and reuse them across session-summary illustrations, NPC reaction images, and end-of-arc trailers in /workflows/multi-shot-short-film.
Is the AI output engine-ready for a 2D rig?
No. AI character output is concept stage — it accelerates the bible build but does not produce engine-ready sprites, normal maps, UV unwraps, or rigged skeletons. For shipping titles, expect a character artist to clean and finalize the AI concept in Procreate or Photoshop, then hand to the rig artist for Spine, Live2D, or DragonBones rigging. Frame Astorie as concept acceleration, not artist replacement.
Generate it on the canvas
Open Astorie, drop this generator on the canvas, and wire it into the workflow you actually need. Free to start — no card required.