Image, video and audio. Each model runs the workflow its authors publish, on a GPU class that fits it, inside your own RunPod account. Three are on when you install; the rest are one switch away.
Z-Image Turbo, Qwen Image 2.1, MiniMax H3. A fresh install starts with these three; their endpoints are created during setup.
One switch in Settings → Models. Turning a model on creates its own endpoint in your RunPod account and pulls its worker image once; turning it off leaves the endpoint idle at no cost.
How turning a model on worksRunPod bills per second of GPU time in your own account; nothing runs while idle.
Each model keeps at most one worker, which stops ten seconds after the last job.
5 models, 3 tools. Create, edit and upscale pictures.
Fast text-to-image: about five seconds a picture on a warm worker, good for iterating on a prompt.
Reference-based image editing with up to ten input images, text-to-image, and a photo restore upscale.
A turbo text-to-image model with a photographic bent and five built-in realism and style LoRAs.
Text-to-image with a Turbo / Default / Quality dial and real guidance control, for prompts that have to be followed closely.
A clean 4x super-resolution to 2K or 4K for pictures that are already sharp.
4 models, 6 tools. Clips from text, stills, keyframes, references or a motion video, with sound.
Open-weight video with sound: five to fifteen second clips from text, a still, two keyframes, references or a pinned timeline.
A second take on text-to-video and image-to-video, up to 2K, with audio and a negative prompt.
Transfers the motion of a video onto a character image, with pose tracking and a choice of background.
Motion control with a face-strength control and relighting, for when the character's face has to carry the performance.
2 models, 3 tools. Voices and music.
Text to speech, voice design from a description, and voice cloning from a few seconds of audio.
Full-band music from a style prompt and optional lyrics, with tempo, key and time signature under your control.
Estimates come from the app's signed catalog: the first pull is the worker image RunPod downloads to a fresh endpoint (10 to 30 minutes, once per endpoint), after which a warm worker answers in seconds. Licenses are the models' own; where a model card is linked, read it before commercial use. Model names belong to their owners.