AI & Workflows

Skytells has been building AI infrastructure since 2012. This section covers the full AI surface in the Console — models, inference, predictions, playgrounds, and Eve.

Skytells has been building AI infrastructure since 2012 — well before generative AI became mainstream. What started as internal machine learning tooling evolved into a full-stack AI platform: proprietary model families, a multi-provider inference layer, developer SDKs, and the Console interface you are using today.

The AI & Workflows section of the Console is the operational surface for everything AI on Skytells: discovering models, running inference, reviewing results, and interacting with Eve — your in-console AI assistant.


Skytells AI since 2012

Skytells entered the AI space as an infrastructure company. Over the years, the platform expanded from internal ML pipelines into a production inference platform with:

  • Proprietary model research — the TrueFusion image generation family, BeatFusion audio models, Mera video models, and more, developed and maintained by Skytells.
  • Multi-provider inference routing — a unified API surface that runs models from OpenAI, Google, Nvidia, Black Forest Labs, and others alongside Skytells' own, with consistent authentication, billing, and response formats.
  • A developer-first API — the Skytells API exposes a Predictions endpoint for generative models and an OpenAI-compatible Inference endpoint for LLMs, with official TypeScript and Python SDKs.
  • The Console — a real-time interface for browsing models, running playground inference, reviewing prediction history, and delegating tasks to Eve.

What Skytells AI provides

Model families

Skytells develops and maintains its own model families across multiple modalities:

Image Generation

The TrueFusion family covers the full quality-cost spectrum — from TrueFusion Edge ($0.01/image, speed-optimized) to TrueFusion Ultra ($0.15/image, enterprise fidelity). Includes variants for panoramic output, style variation, and image-to-image editing.

Video Generation

TrueFusion Video and TrueFusion Video Pro generate video from text or image input. Mera is Skytells' high-fidelity video model for cinematic output.

Audio & Music

BeatFusion generates full music tracks from text descriptions. Audio models cover speech synthesis, sound design, and music composition.

Text & LLMs

Skytells hosts and routes large language models including OpenAI's GPT family, Codex, and instruction-following models — all accessible through the OpenAI-compatible Inference API.

Embeddings

Vector embedding models for semantic search, classification, and retrieval-augmented generation pipelines.

Third-party models

A curated set of third-party models from Black Forest Labs (Flux 2), Google (Imagen 3), Nvidia (Sana), and OpenAI — unified under the same API key, billing system, and Console interface.

For the full model list with namespaces, pricing, and schemas, see the Model Catalog API and List Models.

Inference infrastructure

Every model on Skytells is served through one of two API patterns:

APIEndpointBest for
Predictions APIPOST /v1/predictionsImage, video, audio, music — async generation with polling or webhooks.
Inference APIPOST /v1/chat/completionsText, LLMs, code, chat — synchronous, OpenAI-compatible.

Both share the same authentication, rate limits, and billing. See the API reference for full documentation.

Guarantees

Zero Data Retention

Selected models offer a zero data retention guarantee — your inputs are not stored on the provider's infrastructure after inference completes.

No Prompt Training

Models with the No Prompt Training guarantee do not use your inputs to fine-tune or train any model, whether Skytells' own or a third-party provider's.

Both guarantees are visible per-model in the Model Catalog column view.


What's in this section

How is this guide?

On this page