Open source models on UGESI

17 engines covering video, utility, image, audio. Every rate is published and shown again before you generate.

AnimateDiff

Classic image animation

4 cr / s

CLIP Interrogator

Classic image to prompt

1 cr / image

CodeFormer

Controllable face fidelity

1 cr / image

ControlNet

Pose and edge control

2 cr / image

dots.ocr

Document text extraction

1 cr / page

FramePack

Long clips on modest compute

4 cr / s

IndexTTS 2

Natural prosody

3 cr / 1k chars

InstantID

Identity from one reference

2 cr / image

IP-Adapter

Style transfer from a reference

2 cr / image

LLaVA

Vision language description

1 cr / image

MMAudio

Sound generated against video

5 cr / clip

Real-ESRGAN

Classic upscaler

2 cr / image

Rembg

Fast background removal

1 cr / image

Robust Video Matting

Frame by frame matting

5 cr / s

RVC Voice Conversion

Voice conversion

4 cr / 1k chars

SwinIR

Classic restoration upscaler

1 cr / image

VideoLLaMA 3

Clip level understanding

7 cr / video

What this lab is known for

Open source has 17 models here, covering video, utility, image, audio.

Rates run from 1 to 7 credits. That is the spread between a draft and a flagship.

Every one of them runs on the same balance as models from other labs. You are not locked into a vendor, and switching costs nothing.

Why the lab matters less than the job

Most people arrive here searching for a lab name. The job decides the result. A portrait needs different strengths than a product shot.

Every model below runs on the same balance. You switch between them in one tool. The rate is confirmed before each run.