Skip to content

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.

fix(panel/image-studio): add flux2_edit_inline + flux2_edit_high_inline

mrbizarro701b7fe23bbd
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.py
index 4c172a9..3ec0597 100644
--- a/mlx_ltx_panel.py
+++ b/mlx_ltx_panel.py
@@ -5352,6 +5352,40 @@ class Handler(BaseHTTPRequestHandler):
mflux_family="flux2",
mflux_quantize=4,
)
+ elif engine_override == "flux2_edit_inline":
+ # FLUX.2 Klein-Edit (distilled klein-4b). Multi-ref via
+ # mflux-generate-flux2-edit's --image-paths (REQUIRED).
+ # ~30 s/image at Q4 + 4 steps. Use this when the user
+ # has a character ref and wants quick iteration —
+ # without it the panel UI route silently dropped refs
+ # because flux2_inline maps to t2i mflux-generate-flux2.
+ # mflux's flux2_edit_generate.py FORCES guidance == 1.0
+ # on distilled bases.
+ cfg = agent_image_engine.ImageEngineConfig(
+ kind="mflux",
+ mflux_model="flux2-klein-4b",
+ mflux_family="flux2_edit",
+ mflux_quantize=4,
+ mflux_steps=4,
+ mflux_guidance=1.0,
+ )
+ elif engine_override == "flux2_edit_high_inline":
+ # FLUX.2 Klein-Base-Edit (NON-distilled klein-base-4b).
+ # Real CFG, ~3-5 min/image at Q8 + 25 steps. The
+ # photoreal path when the user has a character ref —
+ # produces photographic output (vs the illustrative
+ # look of distilled flux2_edit). Pass the SHORTHAND
+ # model name; mflux's distilled-detection looks for
+ # "base" substring in the resolved config name and
+ # rejects guidance != 1.0 if missing.
+ cfg = agent_image_engine.ImageEngineConfig(
+ kind="mflux",
+ mflux_model="flux2-klein-base-4b",
+ mflux_family="flux2_edit",
+ mflux_quantize=8,
+ mflux_steps=25,
+ mflux_guidance=4.0,
+ )
elif engine_override == "z_image_turbo_inline":
cfg = agent_image_engine.ImageEngineConfig(
kind="mflux",
@@ -12692,8 +12726,10 @@ HTML = r"""<!doctype html>
<option value="qwen_edit_inline">Qwen-Image-Edit-2509 (multi-ref · 8-step Q4, ~1 min/image)</option>
<option value="qwen_edit_lightning_inline">Qwen-Image-Edit-2509 + Lightning (multi-ref · 4-step Q4, ~10-15 s/image)</option>
<option value="qwen_edit_high_inline">Qwen-Image-Edit-2509 high quality (multi-ref · 30-step Q8, ~5 min/image)</option>
- <option value="flux2_inline">FLUX.2 [klein] 4B (fast T2I)</option>
- <option value="z_image_turbo_inline">Z-Image-Turbo (compact T2I)</option>
+ <option value="flux2_edit_high_inline">FLUX.2 Klein-Base-Edit photoreal (multi-ref &middot; 25-step Q8, ~3-5 min/image)</option>
+ <option value="flux2_edit_inline">FLUX.2 Klein-Edit fast (multi-ref &middot; 4-step Q4, ~30 s/image)</option>
+ <option value="flux2_inline">FLUX.2 [klein] 4B (fast T2I, no refs)</option>
+ <option value="z_image_turbo_inline">Z-Image-Turbo (compact T2I, no refs)</option>
<option value="mock_inline" id="imgStudioMockOption" hidden>Mock (testing — debug only)</option>
</select>
</div>