feat(ltx): wire last-frame conditioning into LTX tab

The LTX tab already collected a last-frame image but anchored every condition at index 0, so it never acted as a last frame. Build a separate condition for it at the final frame: index -1 for the 2.x family (latent index, negatives wrap) and num_frames-1 for 0.9 (pixel index). The Last image input now shows only for Condition models, the pipelines that accept multi-frame conditioning.

Author: CalamitousFelicitousness

modules/ltx/ltx_process.py

@@ -234,8 +234,6 @@ def abort(e, ok: bool = False, p=None):
         condition_images = []
         if ltx_init_image is not None:
             condition_images.append(ltx_init_image)
-        if condition_last is not None:
-            condition_images.append(condition_last)
         conditions = []
         conditions_stage2 = []
         if caps.supports_multi_condition:
@@ -246,14 +244,14 @@ def abort(e, ok: bool = False, p=None):
                 base_w, base_h, condition_strength,
                 condition_images, condition_files, condition_video,
                 condition_video_frames, condition_video_skip,
-                family=caps.family,
+                family=caps.family, num_frames=get_frames(frames), condition_last=condition_last,
             )
             if (final_w, final_h) != (base_w, base_h):
                 conditions_stage2 = get_conditions(
                     final_w, final_h, condition_strength,
                     condition_images, condition_files, condition_video,
                     condition_video_frames, condition_video_skip,
-                    family=caps.family,
+                    family=caps.family, num_frames=get_frames(frames), condition_last=condition_last,
                 )
             else:
                 conditions_stage2 = conditions

modules/ltx/ltx_ui.py

@@ -16,6 +16,7 @@ def _model_change(model_name: str):
         return (
             gr.update(visible=False),   # input_media_accordion
             gr.update(visible=False),   # multi_condition_group
+            gr.update(visible=False),   # last_image
             gr.update(visible=False),   # upsample_accordion
             gr.update(visible=False),   # refine_accordion
             gr.update(value=False),     # upsample_enable (reset)
@@ -38,6 +39,7 @@ def _model_change(model_name: str):
     return (
         gr.update(visible=caps.supports_input_media),
         gr.update(visible=caps.supports_multi_condition),
+        gr.update(visible=caps.supports_multi_condition),  # last_image
         gr.update(visible=True),
         gr.update(visible=True),
         gr.update(value=False),
@@ -73,7 +75,7 @@ def create_ui(prompt, negative, styles, overrides, mp4_fps, mp4_interpolate, mp4
                 ltx_init_image = gr.Image(label='Image', elem_id='ltx_init_image', type='pil', image_mode='RGB', width=256, height=256)
                 ltx_condition_strength = gr.Slider(label='LTX input strength', minimum=0.0, maximum=1.0, step=0.05, value=1.0, elem_id='ltx_condition_strength')
                 with gr.Row():
-                    last_image = gr.Image(label='Last image', elem_id='ltx_last_image', type='pil', image_mode='RGB', width=256, height=256)
+                    last_image = gr.Image(label='Last image', elem_id='ltx_last_image', type='pil', image_mode='RGB', width=256, height=256, visible=False)
                 multi_condition_group = gr.Group(visible=False)
                 with multi_condition_group:
                     gr.Markdown('**Prefix conditioning**: supply a video or gallery to anchor the opening frames', elem_id='ltx_prefix_conditioning_label')
@@ -123,6 +125,7 @@ def create_ui(prompt, negative, styles, overrides, mp4_fps, mp4_interpolate, mp4
         outputs=[
             input_media_accordion,
             multi_condition_group,
+            last_image,
             upsample_accordion,
             refine_accordion,
             upsample_enable,

modules/ltx/ltx_util.py

@@ -138,15 +138,15 @@ def _condition_cls(family: str):
     return LTXVideoCondition
 
 
-def make_condition(condition_cls, family: str, frames, strength: float, is_video: bool):
+def make_condition(condition_cls, family: str, frames, strength: float, is_video: bool, index: int = 0):
     if family == '2.x':
-        return condition_cls(frames=frames, index=0, strength=strength)
+        return condition_cls(frames=frames, index=index, strength=strength)
     if is_video:
-        return condition_cls(video=frames, frame_index=0, strength=strength)
-    return condition_cls(image=frames, frame_index=0, strength=strength)
+        return condition_cls(video=frames, frame_index=index, strength=strength)
+    return condition_cls(image=frames, frame_index=index, strength=strength)
 
 
-def get_conditions(width, height, condition_strength, condition_images, condition_files, condition_video, condition_video_frames, condition_video_skip, family: str = '0.9'):
+def get_conditions(width, height, condition_strength, condition_images, condition_files, condition_video, condition_video_frames, condition_video_skip, family: str = '0.9', num_frames=None, condition_last=None):
     condition_cls = _condition_cls(family)
     if condition_cls is None:
         return []
@@ -186,6 +186,18 @@ def get_conditions(width, height, condition_strength, condition_images, conditio
                 log.debug(f'Video condition: family={family} frames={len(condition_frames)} size={condition_frames[0].size} strength={condition_strength}')
         except Exception as e:
             log.error(f'LTX condition video: {e}')
+    if condition_last is not None:
+        try:
+            if isinstance(condition_last, str):
+                from modules.api.api import decode_base64_to_image
+                condition_last = decode_base64_to_image(condition_last)
+            condition_last = condition_last.convert('RGB').resize((width, height), resample=Image.Resampling.LANCZOS)
+            # 2.x reads index as a latent index and accepts -1 for the final frame; 0.9 uses a pixel index.
+            last_index = -1 if family == '2.x' else max((num_frames or 1) - 1, 0)
+            conditions.append(make_condition(condition_cls, family, condition_last, condition_strength, is_video=False, index=last_index))
+            log.debug(f'Video condition: family={family} last={condition_last.size} index={last_index} strength={condition_strength}')
+        except Exception as e:
+            log.error(f'LTX condition last image: {e}')
     return conditions