Rework vision encoding prompts for proper image framing

Sampler vision re-encoding: prefix scene text with
"Continuing from this scene: {text}" so the model treats the
previous frame as temporal context, not an unrelated subject.

Studio per-scene reference conditioning: re-enable reference_image
with prefix "The character shown performs the following: {encode_text}"
so the model knows the image is a character reference, not something
to describe. Only applies when source_image is connected.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

conditioning.py

@@ -6397,7 +6397,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
             scene_effects.append(effect)
         return scene_texts, scene_effects
 
-    def _v2_transition_scene_conditionings(self, clip, scene_texts, scene_effects=None, encode_cache=None):
+    def _v2_transition_scene_conditionings(self, clip, scene_texts, scene_effects=None, encode_cache=None, reference_image=None):
         if clip is None:
             return None
         scene_texts = list(scene_texts or [])
@@ -6412,7 +6412,8 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
             seen_count = seen_texts.get(scene_text, 0)
             seen_texts[scene_text] = seen_count + 1
             encode_text = f"Returning to an earlier scene: {scene_text}" if seen_count > 0 else scene_text
-            cond, meta, _ = self._v2_encode_prompt(clip, encode_text, encode_cache=encode_cache)
+            vision_text = f"The character shown performs the following: {encode_text}" if reference_image is not None else encode_text
+            cond, meta, _ = self._v2_encode_prompt(clip, vision_text, encode_cache=encode_cache, reference_image=reference_image)
             if not isinstance(cond, torch.Tensor):
                 return None
             scene_meta = dict(meta) if isinstance(meta, dict) else {"pooled_output": None}
@@ -11839,6 +11840,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
                 if scene_texts:
                     scene_conditionings = self._v2_transition_scene_conditionings(
                         clip, scene_texts, scene_effects=split_scene_effects, encode_cache=encode_cache,
+                        reference_image=source_image,
                     )
                     if scene_conditionings is not None:
                         output_conditioning = self._v2_apply_scene_seed_metadata(

samplers.py

@@ -1366,7 +1366,7 @@ class FunPackLTXAVSceneChainSampler:
             text = existing_meta.get("funpack_encode_text") or self._scene_text(next_cond, next_index)
             if not text:
                 return None
-            tokens = clip.tokenize(text, image=pixel_frame, skip_template=False)
+            tokens = clip.tokenize(f"Continuing from this scene: {text}", image=pixel_frame, skip_template=False)
             encoded = clip.encode_from_tokens_scheduled(tokens)
             if not isinstance(encoded, list) or not encoded:
                 return None