Polyfaced

About Polyfaced — Multi-Model AI Creation Platform

Polyfaced is a multi-model AI creation platform — Kling 2.1 video, GPT Image 2 image, more coming. Our mission: ship pro AI assets without juggling 5 vendor tabs.

Our mission

Polyfaced exists because generative AI is moving faster than any single provider, but creators are still stuck juggling separate accounts, separate billing pages, and separate prompt boxes across Kuaishou, OpenAI, RunwayML, Google, Stability and a dozen smaller labs. The state-of-the-art video model lives on one platform. The state-of-the-art image model lives on another. The state-of-the-art music model is somewhere else entirely. You sign up everywhere, learn five different parameter schemas, top up five different credit balances, and lose two hours every week to context-switching.

We think a creator should be able to open one studio, pick the right model for the job, generate the asset, and ship — without ever caring which lab trained the model or which infrastructure ran the inference. That's the platform we're building.

What v1 delivers

Polyfaced v1 launches with the two models we consider the practical default for generative video and generative image in 2026:

  • Kling 2.1 (Kuaishou) for video — competitive with Sora 2 and Veo 3 on most benchmark sets, especially strong on motion physics, character consistency, and product shots. Up to 10 seconds at 1080p, both text-to-video and image-to-video, three aspect ratios out of the box.
  • GPT Image 2 (OpenAI) for image — ahead of peer models on prompt faithfulness, native typography, and image-to-image with subject preservation. Up to 1536px, five aspect ratios, both modalities.

Both models share a unified prompt box, a unified credit ledger, and unified Cloudflace R2 storage with permanent share URLs. Three more surfaces (AI Avatar, AI Music, Movie Maker) are publicly on the waitlist and will land as their own dedicated tool pages.

How we think about pricing

The credit ledger is FIFO and refundable. Every generation freezes credits up front, settles on success, releases on failure — so you never pay for an output you didn't receive. Free plan ships with 5 sign-up credits, enough to evaluate both models before committing. Pro plan is $29.9/mo for 800 credits with no annual lock and full commercial license. Credit packs from $4.99/100 cover the gap when monthly credits run low. Cancel anytime, no contracts, no hidden cancellation fees.

How we think about content policy

Polyfaced inherits the content policy of the underlying provider — Kuaishou for Kling 2.1, OpenAI for GPT Image 2. Content policy violations on the provider side trigger a failed generation and an automatic credit refund; we don't moderate independently on top. The policies in plain language: no extremist content, no impersonation of real public figures without context, no copyright-circumventing reproductions of branded characters. If you're unsure whether a creative direction is allowed, the safe bet is to ask the provider directly via their published policy pages — we link them in the per-model FAQ on each studio page.

The roadmap

The short list, in priority order:

  1. Avatar surface — text-to-avatar and image-to-avatar with consistency across a batch. Targeting Q3 2026.
  2. Music surface — text-to-song with vocal direction, genre control, and stem export. Targeting later in 2026.
  3. Movie Maker — multi-shot composition with character continuity and audio. Targeting Q4 2026.
  4. Multi-model selector inside Studio — switch between Kling 2.1, Veo 3.1, Sora 2, Seedance, Wan 2.5, Pika 2.2 from the same prompt box. v2.
  5. Public API — for builders integrating Polyfaced as a backend. Targeting v2 once the creator experience hardens.

We don't ship roadmap items that don't deliver real value at launch — Avatar and Music will not land as half-working previews. When you see them appear in the navigation as something other than "Coming Soon", you'll know they're production-ready.

Get in touch

Polyfaced is built by YDD Creation, an independent studio focused on creator-facing AI tools. We are not affiliated with Kuaishou or OpenAI; we run their models under their official terms via the Kie infrastructure layer.