Free, open-source AI video generator for Mac. Run LTX-2.5 and MiniMax H3 (Hailuo) locally on Apple Silicon: text-to-video and image-to-video with sound, lip-sync, character LoRA training, songs from lyrics (YuE2) and image editing (Qwen-Image-Edit). One-click install via Pinokio. Offline: no cloud, no API key, no NVIDIA GPU.
Python ai-filmmakingai-videoai-video-generatorapple-siliconhailuoimage-to-videolip-synclocal-ailora-trainingltx-2ltx-2-5ltx-videomacosminimax-h3mlxmusic-generationoffline-aipinokiotext-to-videovideo-generation
fix(agent): default Qwen-Edit presets to 2511 (better consistency)
mrbizarro7dbbe3c6090a
diff --git a/agent/tools.py b/agent/tools.pyindex 53e91a2..aa977b4 100644--- a/agent/tools.py+++ b/agent/tools.py@@ -1168,30 +1168,48 @@ def _generate_shot_images(args: dict, ops: PanelOps, session: dict) -> dict:mflux_guidance=4.0,)elif engine_override == "qwen_edit_lightning":+ # Qwen-Image-Edit-2511 + Lightning LoRA 4-step. 2511 is the+ # current best Qwen-Edit (Apache 2.0) — significantly stronger+ # character consistency than 2509, mitigated image drift on+ # multi-ref edits, multi-person fusion, built-in popular LoRAs.+ # The Lightning LoRA filename pin avoids mflux downloading all+ # 4 variants (bf16/fp32 × 4-step/8-step) and getting stuck on+ # ambiguous resolution.cfg = _image_engine.ImageEngineConfig(kind="mflux",- mflux_model="Qwen/Qwen-Image-Edit-2509",+ mflux_model="Qwen/Qwen-Image-Edit-2511",mflux_family="qwen_edit",- mflux_quantize=4,+ mflux_quantize=6,mflux_steps=4,- mflux_lora_paths=["lightx2v/Qwen-Image-Edit-2511-Lightning"],+ mflux_lora_paths=[+ "lightx2v/Qwen-Image-Edit-2511-Lightning:Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors"+ ],mflux_lora_scales=[1.0],)elif engine_override == "qwen_edit":+ # Standard Qwen-Edit-2511 at Q6 8-step. 2511 base, no+ # Lightning. ~1 min/image on M4 Max. Iteration tier with+ # 2511's improved consistency over 2509.cfg = _image_engine.ImageEngineConfig(kind="mflux",- mflux_model="Qwen/Qwen-Image-Edit-2509",+ mflux_model="Qwen/Qwen-Image-Edit-2511",mflux_family="qwen_edit",- mflux_quantize=4,+ mflux_quantize=6,mflux_steps=8,)elif engine_override == "qwen_edit_high":+ # Final-render preset: Qwen-Image-Edit-2511 Q8 + 40 steps.+ # 40 steps matches the Diffusers reference example for+ # Qwen-Edit (qwen.ai blog: 2511 official inference uses+ # num_inference_steps=40 + true_cfg_scale=4.0 ++ # guidance_scale=1.0 + negative_prompt=" "). mflux's --guidance+ # maps to true_cfg_scale on the qwen-edit CLI.cfg = _image_engine.ImageEngineConfig(kind="mflux",- mflux_model="Qwen/Qwen-Image-Edit-2509",+ mflux_model="Qwen/Qwen-Image-Edit-2511",mflux_family="qwen_edit",mflux_quantize=8,- mflux_steps=30,+ mflux_steps=40,mflux_guidance=4.0,)elif engine_override == "kontext_high":