Connecting AI, humans, and the physical world. TTS｜ASR｜VL ｜Audio｜Video｜Contextual Intelligence ⚡️Vibe with us: https://xitter.cf/t.co/Lmk4Kd65JJ
Joined January 2026
- Tweets367
- Following28
- Followers2.3K
- Likes336
Pinned Tweet
MOSS-Image API is coming soon in 24hrs
Join MOSS discord for FREE early access 👉 discord.gg/Wjku5AYyxx
Upload your dataset and train a MOSS-Image-SFT LoRA on @ModelScope2022 .
Thanks to the ModelScope team for bringing MOSS-Image-SFT to their training platform!
Start training ➡️ modelscope.cn/aigc/model-tra…
8-step generation with the MOSS-Image-SFT Turbo LoRA. ⚡
Results are close to 50-step MOSS-Image-SFT inference in the shared examples. Trained using DiffSynth-Studio's OPD (On-Policy Distillation).
Try it ➡️ modelscope.cn/models/DiffSyn…
Faster MOSS-Image inference with EntroPack INT8 + DiffSynth-Studio pipeline compilation. ⚡
In the reported test: 974 → 507 ms/step.
Inference script ➡️ github.com/modelscope/DiffSy…
Local inference & training ➡️ diffsynth-studio-doc.readthe…
8-step generation with the MOSS-Image-SFT Turbo LoRA. ⚡
Results are close to 50-step MOSS-Image-SFT inference in the shared examples. Trained using DiffSynth-Studio's OPD (On-Policy Distillation).
Try it ➡️ modelscope.cn/models/DiffSyn…
Nearly 2× faster per inference step in the reported test, using EntroPack's INT8 backend + DiffSynth-Studio's pipeline compilation. ⚡
Inference script ➡️ github.com/modelscope/DiffSy…
Local inference & training docs ➡️ diffsynth-studio-doc.readthe…
MOSS retweeted
MOSS Image dropped less than a week ago and it’s already at 1.29M downloads on Hugging Face.
If you have consumer hardware, this open source image model is for you + FREE workflow below
→ 5B params
→ int8 weights are only 5.4GB
→ Apache 2.0 license
→ seamless install via Comfy Desktop
The quality punches way above it's weight. Great work by @Open_MOSS.
What the ComfyUI community shipped in 5 days:
→ color + depth control adapter (darask0)
→ hires LoRA, native 4096×2304 gens (thee comfyanonymous)
→ LoRA training support all the major tools
→ Kromoss finetune, with a 8step DMD2 turbo LoRA (lodestones)
→ DSLR realism LoRA (Danrisi)
I built a workflow for the color-depth control adapter running in ComfyUI. Prompt adherence and detail is spot on (check the lens flare).
MOSS retweeted
MOSS image FAST AS HELL TURBO by @_akhaliq
50 steps → 4 steps ⚡
turbo feels like a sweet spot for generating lots of images fast
regular text posters still look pretty good too. tiny lettering gets a bit blurry, so i'd stick with the original SFT for text-heavy designs!
Base 50 steps 🆚 SFT 50 steps 🆚 4-step version side by side comparison👇
MOSS retweeted
MOSS-Image SFT and MOSS-Image 8-Step are now on Sogni, and both are included in Sogni Unlimited
It’s OpenMOSS’s 5-billion-parameter text-to-image model, made for text. Quote the words in your prompt and it writes them out exactly, from long instructions and terminal commands to full menus, in English and Chinese.
Every image in this thread was made with it 👇🧵
MOSS retweeted
MOSS Image is live on Civitai.
A 5B open-weights text-to-image model from the OpenMOSS team. Text that reads right, JSON layout control, and prompts up to 4K tokens. Apache 2.0.
You can also train LoRAs on it in the Training Studio.
Try it: civitai.com/articles/36843
Try MOSS-Image in ComfyUI with our free API. 🎨
⚡ Get your API key: platform.mosi.cn/?utm_source…
Love seeing what you're building with MOSS-Image. Thanks for sharing your work.
If you'd like to experiment too, our MOSS-Image API is free to try: platform.mosi.cn/?utm_source…
Here are a few community projects and experiments that caught our eye 👇
16×16-style game icons with a pixel-art LoRA.
xitter.cf/Angaisb_/status/211121…
Manga panels with vertical Japanese text and sound effects, using layout control and a character LoRA.
xitter.cf/sep_is_heim/status/211…
