Getting started

Models & libraries

Friends & Robots draws on three resources: hosted generative models for creation, tooling bundled with the app for precise, deterministic work, and licensed stock for ready-made media. The agent reaches for whichever fits the step.

Generative models#

The agent knows each model's strengths and weaknesses and reaches for the right one for the task — so you don't have to choose. You can still call out a specific model whenever you want.

This list is curated and fluid. We continuously sample new models and expose only the best in class — when a stronger one emerges, it takes the slot.

CapabilityModels & providers
Image generation FLUX.2 (max, pro, flex) Recraft v4.1 Seedream v5.0 GPT Image 2 Nano Banana Pro Grok Imagine Quality
Video generation Kling v3 Omni Seedance 2.0 Happy Horse 1.0
Enhancement & restoreTopaz Labs (10+ models)
Speech & voiceEleven V3
Music Gemini Lyria 3 Pro Eleven Music V2
TranscriptionEleven Scribe V2

Bundled libraries#

These ship inside the desktop app and run locally in your session folder — sandboxed, with no internet access. They handle conversions, compositing, and analysis deterministically, so the same input always yields the same output. Nothing to install; the agent runs them for you.

FunctionPowered by
Video & audio convert, trim, compress, GIF

ffmpeg

Image convert, resize, composite, sheets

ImageMagick

SVG → raster rendering

resvg

Image & audio analysis and processing

OpenCV, NumPy, Pillow, librosa

Timeline compositing — captions, kinetic text, audiograms

Remotion

HTML → PDF

Chromium

Stock media#

Licensed stock the agent can search and pull from, so real footage, photos, and audio drop straight into a project without leaving the conversation.

TypeSources
Images Pexels Storyblocks
VideoStoryblocks
Sound effectsStoryblocks
MusicStoryblocks