---
title: sensenova-u1.5-preview
canonical_url: "https://www.modelscope.cn/studios/SenseNova/sensenova-u1.5-preview"
md_url: "https://www.modelscope.cn/studios/SenseNova/sensenova-u1.5-preview.md"
repository: SenseNova/sensenova-u1.5-preview
chinese_name: "SenseNova U1.5 Preview"
last_updated: 2026-08-05
sdk_type: gradio
sdk_version: 6.17.3
downloads: 0
stars: 58
---

# sensenova-u1.5-preview

> sensenova-u1.5-preview - SenseNova 在 ModelScope 创建的在线 Demo。Unified image generation & editing with SenseNova/SenseNova-U1.5-8B-MoT-Preview (NEO-Unify).

SenseNova/sensenova-u1.5-preview 是 ModelScope 魔搭社区上的在线可交互 Demo（创空间），基于 gradio 6.17.3 构建，中文名为「SenseNova U1.5 Preview」。

- **Repository**: SenseNova/sensenova-u1.5-preview
- **SDK**: gradio
- **SDK version**: 6.17.3
- **Downloads**: 0
- **Stars**: 58
- **Last updated**: 2026-08-05

Source: https://www.modelscope.cn/studios/SenseNova/sensenova-u1.5-preview

---

# SenseNova-U1.5-8B-MoT-Preview 📊

Unified **image generation and editing** with
[`SenseNova/SenseNova-U1.5-8B-MoT-Preview`](https://modelscope.cn/models/SenseNova/SenseNova-U1.5-8B-MoT-Preview),
running on ZeroGPU.

SenseNova-U1.5-8B-MoT-Preview is a monolithic **any-to-any (NEO-Unify)** model that
unifies multimodal understanding, generation, and editing in one architecture (no separate
visual encoder / VAE). It renders complex, text-dense infographics (charts, posters, knowledge
illustrations) across English and Chinese, and edits existing images via
natural-language instructions: local text edits, content insert / remove / replace, global
style changes, and layout tweaks.

This Space exposes two tabs:

- **Generate** — text → infographic (`t2i_generate`).
- **Edit** — image + instruction → edited image (`it2i_generate`).

Generation and editing use the model's officially recommended settings:
`cfg_scale=4.0`, `img_cfg_scale=1.0`, `cfg_norm="global"`, `timestep_shift=3.0`, `num_steps=50`.

The `sensenova_u1` package is vendored locally to register the NEO-Unify model
architecture with `transformers`.