Skip to content
Agientry
← Back to Videos

Qwen-Image-2.1-Turbo Converted to Apple Core AI for On-Device RGBA Generation

Loading video
Loading video

A conversion of Qwen-Image-2.1-Turbo, the 8-step version of the Qwen team 7B text-to-image model, into Core AI, the on-device runtime from Apple. It can make transparent images with a real alpha channel: on an M4 Max a 512x512 sticker took 9.7 seconds, timed by the app, with the converted weights published on Hugging Face.

Qwen-ImageCore AIApple SiliconOn-Device InferenceText-to-ImageRGBATransparent ImageM4 MaxHugging Face
Category: image-generation
Author: @JackdeS11
Date: 2026-10-11T00:00:00
Duration: 14.9s
Source: https://x.com/JackdeS11/status/2109325759915737244
Reference: https://huggingface.co/mlboydaisuke/RGBA-Image-2.1-Turbo-CoreAI

Related Videos

Qwen-Image-2.1-Turbo lands on HuggingFace: 8 steps, same quality, 4x faster
0:21

Qwen-Image-2.1-Turbo lands on HuggingFace: 8 steps, same quality, 4x faster

@HuggingApps · 179 views · 2026-10-09
Qwen-ImageText-to-ImageDiffusion Distillation
Qwen-Image-2.1 consistent-lighting LoRA: blend objects into scenes with natural shadows, no glare
0:10

Qwen-Image-2.1 consistent-lighting LoRA: blend objects into scenes with natural shadows, no glare

@wildmindai · 1k views · 2026-10-04
Qwen-ImageLoRALighting
Qwen-Image-2.1 outfit-swap consistency LoRA: dress a person from two pics with 0.1px drift
0:19

Qwen-Image-2.1 outfit-swap consistency LoRA: dress a person from two pics with 0.1px drift

@wildmindai · 422 views · 2026-10-04
Qwen-ImageLoRAOutfit-Swap
Qwen-Image-2.1-Fix: one LoRA repairs Qwen Image 2.1's common generation flaws without touching the base checkpoint
0:27

Qwen-Image-2.1-Fix: one LoRA repairs Qwen Image 2.1's common generation flaws without touching the base checkpoint

@Oluwaphilemon1 · 588 views · 2026-10-01
Image GenerationQwen Image 2.1LoRA
Ideogram 4.5 live on fal: ten consecutive edits with 94–99% of unasked pixels unchanged, typography rendered exactly in any language
0:16

Ideogram 4.5 live on fal: ten consecutive edits with 94–99% of unasked pixels unchanged, typography rendered exactly in any language

@fal · 477 views · 2026-09-30
Image GenerationIdeogramfal
Google AI Edge Foresight: Free Offline AI Note-Taking on Apple Silicon
0:47

Google AI Edge Foresight: Free Offline AI Note-Taking on Apple Silicon

@AYi_AInotes · 89 views · 2026-10-11
Google AI EdgeOn-Device AINote Taking

As an Amazon Associate, we earn from qualifying purchases.