地上最強免費開源A.I video [WAN2.2]

虐待 A.I

Reddit啲 example幾好
https://www.reddit.com/r/comfyui/s/KXhvmFr6wC

AI /AI video 每個月都進化中

有冇玩Qwen Image?
Download左但未試，希望會有Lora 同uncensored checkpoint

未
跟唔到太貼

ice bucket 虐待系列:

Qwen Image 可以出到繁體中文, 支援mutlit language prompt...實測同即夢既質素差唔多 , gpt4o 既prompt adherence.佢個CLIP 真係好勁....
nsfw 只係有時得...似乎原生model 都係有censor, 可以自己train lora 解決

睇黎開源既生態圈有一番惡鬥...

最近出既Flux-Krea 又係好有potential...改善左Flux 既plastic skin , 而且artistic 好多

但係我想玩依個multiTalk先, 自備 audio 做到類似Veo3嘅嘢
https://github.com/MeiGen-AI/MultiTalk

multiTalk 我有用開, 係comfyui 亦都用到, 算係最強既開源avatar..即係你input 一段voice over +prompt 可以叫佢唱歌跳舞, 雙人都可以...

但係唔同VEO3, VEO3 係自帶埋audio generation , dialogue, soundtracks, sound effects

我知，但係暫時最接近Veo3應該係佢

飛釘 bam得

只眼

quality set 低咗

https://www.reddit.com/r/comfyui/comments/1mixibz/flux_krea_nunchaku_vs_wan22_lightxv_lora_using/

Krea vs Wan, 玩圖片display card 6G vram都得