modelscope/DiffSynth-Studio

View on GitHub

Enjoy the magic of Diffusion models!

Stars: 13.1k
Forks: 1.3k
Open beginner issues: 1
Indexed issues: 542
Avg merge: 13h 12m
Merged PRs (30d): 45

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 20, 2026
Latest indexed: Sep 20, 2026
Contributing guide: No contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

542 open issues indexed

About Wan2.2 Vace Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

modelscope/DiffSynth-Studio#1430 · 1 comment · Opened May 6, 2026
Wan2.2-Fun-A14B-Control-Camera.py Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1425 · 1 comment · Opened Apr 30, 2026
RuntimeError: cuFFT doesn't support tensor of type: BFloat16 Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1423 · 1 comment · Opened Apr 29, 2026
首尾帧训练时的loss问题 Open

Difficulty 4/5 3-5 days Newbie friendliness 38/100

modelscope/DiffSynth-Studio#1422 · 2 comments · Opened Apr 29, 2026
modality cfg in ltx-2.3 Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1420 · 2 comments · Opened Apr 29, 2026
关于Z-Image Omni代码中的文本编码长度问题 Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

modelscope/DiffSynth-Studio#1418 · 1 comment · Opened Apr 28, 2026
怎么对qwen-image-edit-2511、flux.2-klein-base-4b等模型做多卡推理？ Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1411 · 3 comments · Opened Apr 24, 2026
ImportError: cannot import name 'SiglipVisionTransformer' from 'transformers.models.siglip.modeling_siglip' Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

modelscope/DiffSynth-Studio#1407 · 1 comment · Opened Apr 23, 2026
Z-Image支持编辑范式训练 Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1405 · 1 comment · Opened Apr 22, 2026
Wan2.2 TI2V首帧出现色差，问题长时间存在 Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1402 · 3 comments · Opened Apr 19, 2026
TypeError: model_fn_joyai_image() missing 3 required positional arguments: 'latents', 'prompt_embeds', and 'prompt_embeds_mask' Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

modelscope/DiffSynth-Studio#1401 · 1 comment · Opened Apr 17, 2026
关于Flux2klein的图像预处理gap Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1398 · 2 comments · Opened Apr 15, 2026
Qwen-Image-Layered Open

Difficulty 4/5 3-5 days Newbie friendliness 28/100

modelscope/DiffSynth-Studio#1396 · 1 comment · Opened Apr 15, 2026
qwen-image-edit-2511部署推理和lora微调 Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1395 · 1 comment · Opened Apr 14, 2026
训练时梯度累积有问题？ Open

Difficulty 3/5 1-2 days Newbie friendliness 55/100

modelscope/DiffSynth-Studio#1388 · 4 comments · Opened Apr 11, 2026
Qwen-Image FP8训练报错 Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

modelscope/DiffSynth-Studio#1386 · 3 comments · Opened Apr 9, 2026
有大佬用wan2.2 Inp 做首尾帧 LORA训练吗？ Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1385 · 1 comment · Opened Apr 9, 2026
wan2.2 如果要训练首尾帧，是不是需要把训练数据变换到81帧以内，不然就被截断了 Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

modelscope/DiffSynth-Studio#1384 · 1 comment · Opened Apr 9, 2026
有人知道wan2.2 VACE 模型，在训练depth control的时候，用的是depth v几的模型吗？另外，depth转rgb灰度图的时候，深的地方是靠近0还是255？ Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

modelscope/DiffSynth-Studio#1381 · 1 comment · Opened Apr 8, 2026
Error loading DinoV3 state dict Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

modelscope/DiffSynth-Studio#1378 · 2 comments · Opened Mar 30, 2026
请问目前只有wan支持多卡推理吗,其他模型都不能多卡推理的. Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1377 · 1 comment · Opened Mar 30, 2026
FLUX2-klein模型多图输入的训练任务 Open

Difficulty 3/5 1-2 days Newbie friendliness 38/100

modelscope/DiffSynth-Studio#1376 · 3 comments · Opened Mar 27, 2026
support daVinci-MagiHuman Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1375 · 1 comment · 2 reactions · Opened Mar 27, 2026
LTX2.3 训练720P 超显存 Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1374 · 2 comments · Opened Mar 26, 2026
Latest update broke Wan2.2 S2V Open

Difficulty 1/5 Under an hour Newbie friendliness 78/100

modelscope/DiffSynth-Studio#1373 · 5 comments · 1 reaction · Opened Mar 25, 2026
About megatron Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1372 · 1 comment · Opened Mar 25, 2026
Wan DiTBlock AutoWrappedNonRecurseModule causes black output for AniSoraV3.2 at 1280x720 on DiffSynth 2.0.6 Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

modelscope/DiffSynth-Studio#1367 · 1 comment · Opened Mar 23, 2026
LTX2.3 data process when input_audio process error Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

modelscope/DiffSynth-Studio#1366 · 1 comment · Opened Mar 23, 2026
wan2.2-I2V-A14B模型如何从已微调的权重中恢复训练？ Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

modelscope/DiffSynth-Studio#1360 · 2 comments · Opened Mar 19, 2026
wan-anmiante多卡分布式训练支持？ Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1359 · 1 comment · Opened Mar 18, 2026
为什么qwen-image系列的inference和training，vae、text_encoder和dit的部分来要配置自于不同模型的权重？ Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1357 · 3 comments · Opened Mar 18, 2026
Question about InputImageEmbedder and EditImageEmbedder in FLUX.2 Open

Difficulty 2/5 1-3 hours Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1356 · 2 comments · Opened Mar 17, 2026
请问能增加ltx2.3图生视频对口型对训练支持吗？ Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1355 · 7 comments · Opened Mar 17, 2026
MOVA模型可以支持T2AV吗？目前代码中一定需要加入Image Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1353 · 3 comments · Opened Mar 16, 2026
LTX2模型可以同时输入首帧参考和视频depth参考吗，即ti2v + depth control? Open

Difficulty 5/5 Over a week Newbie friendliness 38/100

modelscope/DiffSynth-Studio#1352 · 10 comments · Opened Mar 16, 2026
lxt2 模型示例推理脚本报RuntimeError: Error(s) in loading state_dict for LTX2TextEncoder: Missing key(s) in state_dict: "vision... Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1351 · 2 comments · Opened Mar 14, 2026
大佬，后面会有wan系列DMD蒸馏的实现吗 Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1347 · 2 comments · 1 reaction · Opened Mar 12, 2026
z_image和z_image_turbo full fineturn 效果很差，感觉这两个模型是真实人像的特调模型吗??? Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1344 · 1 comment · Opened Mar 11, 2026
Scale Issue about Camera Controlled Video Model Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1340 · 1 comment · Opened Mar 9, 2026
请问有没有办法实现在开启fp8和offload后实现多卡训练wan2.2 animate？ Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1328 · 4 comments · Opened Mar 5, 2026
Add support for Helios-14B, a real-time long video generation model Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1327 · 1 comment · Opened Mar 5, 2026
[FETAURE] Add WAN Vace LoRA train with full dataset Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1326 · 1 comment · 1 reaction · Opened Mar 4, 2026
wan2.2-animate 长视频如何生成？直接进行拼接画面突变差异太大。 Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1316 · 1 comment · Opened Mar 2, 2026
Is there any guidance for super resolution? Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1314 · 1 comment · Opened Feb 28, 2026
FLUX.2-klein模型训练的lora没有效果 Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1313 · 2 comments · Opened Feb 28, 2026
Wan2.2-TI2V-5B lora训练崩溃，全是噪声 Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1311 · 2 comments · Opened Feb 27, 2026
微调vae Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

modelscope/DiffSynth-Studio#1308 · 5 comments · Opened Feb 25, 2026
请教大佬，如何多机训练FLUX.1-Kontext-dev Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

modelscope/DiffSynth-Studio#1303 · 1 comment · Opened Feb 20, 2026
Jointly train LoRA and DiT Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

modelscope/DiffSynth-Studio#1301 · 2 comments · Opened Feb 18, 2026
想问下，如何把image-edit模型的text-encoder，从llm改成vlm，比如qwen3-8b，改成qwen3-vl-8b-instruct？ Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

modelscope/DiffSynth-Studio#1298 · 3 comments · Opened Feb 13, 2026

modelscope/DiffSynth-Studio

Get new issues in your inbox