LFM3-Omni-2B
Any-to-any: text/image/video/audio in, text or speech out
None defined yet.
Any-to-any: text/image/video/audio in, text or speech out
LiquidAI vision-language model for images and video
Document image OCR — text, tables, formulas, seals, charts
Ground referring expressions with draft-and-verify boxes
English ASR with a looped transformer (153M params)
Streaming multilingual TTS with zero-shot voice cloning
Simultaneous Chinese⇄English speech translation
Simultaneous English⇄Chinese speech translation
Agentic GUI grounding that zooms in with crops
Locally controllable 3D generation from primitives
Zero-shot English TTS and voice conversion, 33.6M params
Interactive text + camera controlled video world model
Coarse-mask local editing with dual-path FLUX.2 Klein 9B
Image-to-3D part-level generation with variable granularity
Textureless city mesh to street views and 3D Gaussians
Step into a live, steerable generated world
Image to high-res 3D mesh with Hi3Dv1 refinement
Solve visual jigsaw puzzles via latent visual reasoning
Grounding, detection, OCR, segmentation & matting in one VLM
Detect, ground, count, OCR, segment & matte with one VLM
Efficient real-world single-image dehazing with PGL-Net
3D point cloud semantic segmentation with LitePT
Embodied task planning VLM (27B, image + instruction)
Pixel-space t2i with a region-token budget dial