SenseNova-U1.5 MoT TailSteer

SenseNova-U1.5 MoT TailSteer

Edits an image over several turns, feeding each result back in as the next input while limiting drift. Suits people who want to try step-by-step image editing or study the model.

Made by hugging-apps on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a demo of multi-turn image editing with a SenseNova 8B model tuned by a method called TailSteer. When each edit becomes the input for the next, small errors pile up and the picture drifts. TailSteer is meant to reduce that. The demo sends every result back in and keeps a history of turns.

Who it suits: It suits people who edit a picture in several small steps, and researchers who want to see how a tuned model holds up across repeated edits.

Industry | Who uses it | How they use it | Example
Design | Designers | Refine a picture through a chain of small instructions | Change the lighting, then adjust a detail, then add an object
Marketing | Content teams | Adjust a visual in steps instead of starting over | Insert a product from a reference image into a scene
Machine learning | Researchers | Watch how drift builds up over many turns | Run ten edits in a row and compare with the start
Software | Developers | Read a Gradio app that wraps an image editing model | Open app.py and follow the turn loop

Skip it if you need exact, repeatable edits at full speed, since the demo trades some sampling steps for speed.

Where to use it:

• A browser: open the Space's page on Hugging Face, with nothing to install
• A copy on Windows, macOS or Linux, using Python 3.12
• Tools that speak MCP, since the Space carries an MCP tag, though the source does not describe its tools

Use case | Who | What they get back
Multi-turn editing | Designers | An edited picture after each instruction, with a turn history
Subject insertion | Content teams | An image that adds a subject from a reference picture
Text to image | Anyone curious | A generated picture you can send into the editing loop

How to set it up: The online copy needs nothing. For a copy on your own computer, you need the following.

• Python 3.12 with pip on Windows, macOS or Linux
• The Hugging Face hub client, which the first step installs
• The packages listed in requirements.txt

1. Install the Hugging Face hub client.
2. Download the Space's files into a local folder.
3. Move into that folder.
4. Install the Gradio version the Space pins.
5. Install the remaining packages from requirements.txt.
6. Start app.py.

pip install -U huggingface_hub
hf download hugging-apps/tailsteer-sensenova-u15-mot --repo-type space --local-dir tailsteer-sensenova-u15-mot
cd tailsteer-sensenova-u15-mot
pip install gradio==6.31.0
pip install -r requirements.txt
python app.py

How to use it: Upload an image, give an instruction and keep going, since each result becomes the next turn's input. You can add optional reference images when you want to insert a subject. The same model can also make a picture from text.

Tool | What it does
Multi-turn editing | Applies an instruction to the current image and feeds the result back for the next turn
Reference images | Optional extra pictures used to insert a subject
Text-to-image | Generates a picture that you can send into the editing loop
Turn history | Keeps the earlier turns visible

Example: You upload a photo of an atrium and type "make the light warmer". The Space returns the edited picture. You then type "add a plant near the stairs", and it edits the new picture, with both turns kept in the history.

For newcomers and experts: If you are new, begin with one of the built-in example pictures and make one small edit at a time. Check each result before you ask for the next change.

If you know the field, the demo follows the official sampling settings of CFG scale 4.0 and timestep shift 3.0, and it normalises edit output to 2048 by 2048 pixels. It uses 28 steps by default instead of 50, which is faster but gives the sampler less room. A guard model checks edit prompts on the processor.

Title: Sampling steps, demo against the official setting
This demo: 28 steps
Official default: 50 steps

You upload an image or generate one
You type an edit instruction
The TailSteer model edits the image
The result becomes the next turn's input

Good to know: The example pictures come from the model's own showcase, and the model code is vendored from the SenseNova-U1 project.

• License: apache-2, and the examples and vendored package are credited as Apache-2.0
• Cost to run: no API key or paid service is named, and the online copy runs on zero-a10g hardware at Hugging Face
• Data it reaches: the pictures and prompts you give it, which the online Space handles on Hugging Face's servers, and a guard model screens edit prompts
• Not for: the source lists no exclusions, so use only pictures you have the right to edit

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/hugging-apps/tailsteer-sensenova-u15-mot
• Files: https://huggingface.co/spaces/hugging-apps/tailsteer-sensenova-u15-mot/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.