Prototipal

Product · Soul

Save who. Generate any scene.

Soul is the fastest path to identity-consistent generation. Save 2–8 photos of any character — yourself, a brand spokesperson, a fictional persona — and place them in any scene with one prompt.

0Training time
8Refs per character
Generations

Capabilities

What it does.

06Features
01

No training

Multi-reference inference — no LoRA, no fine-tuning, no compute wait.

02

Reusable library

Save once, generate forever. Characters persist across sessions.

03

Identity locked

Face, body proportions, signature features stay consistent.

04

Any model

Works with Nano Banana 2 (image), Veo 3.1, Kling, Sora 2 (video).

05

Motion preset compatible

Layer cinematic camera moves on every Cast generation.

06

Public / private

Keep personal characters private. Share community casts publicly.

Samples

Generated here.

Every clip below was rendered through Prototipal — no curation, no cherry-picked external content.

06Samples
01
Cast across scenes
02
Sample
Identity preserved
03
Multi-shot consistency
04
Sample
Brand spokesperson
05
Character action
06
Sample
Outfit try-on

FAQ

Common questions.

01Why not LoRA training?

Multi-reference inference is faster (no training step), cheaper (no compute charges), and matches LoRA quality for most use cases. We may add LoRA training later for power users.

02How many references should I upload?

4-6 from different angles works best — front, three-quarter left/right, profile, smiling and neutral. Clean backgrounds preserve identity better.

03Can I share characters?

Yes — toggle the public flag on any character. Future versions will surface a /soul-community feed of shared characters.

Image cast: 2 credits · Video cast: from 4.5 credits/sec

Open the Soul library