Update app.py

app.py

@@ -11,6 +11,7 @@ from transformers import (
     Qwen2_5_VLForConditionalGeneration,
     Qwen2VLForConditionalGeneration,
     Glm4vForConditionalGeneration,
+    AutoModelForVision2Seq,
     AutoProcessor,
     TextIteratorStreamer,
 )
@@ -61,6 +62,15 @@ model_s = Glm4vForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
+# Load kanana-1.5-v-3b-instruct 
+MODEL_ID_F = "kakaocorp/kanana-1.5-v-3b-instruct"
+processor_f = AutoProcessor.from_pretrained(MODEL_ID_F, trust_remote_code=True)
+model_f = AutoModelForVision2Seq.from_pretrained(
+    MODEL_ID_F,
+    trust_remote_code=True,
+    torch_dtype=torch.float16
+).to(device).eval()
+
 def downsample_video(video_path):
     """
     Downsample a video to evenly spaced frames, returning each as a PIL image with its timestamp.
@@ -103,6 +113,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "GLM-4.1V-9B-Thinking":
         processor = processor_s
         model = model_s
+    elif model_name == "kanana-1.5-v-3b":
+        processor = processor_f
+        model = model_f
     else:
         yield "Invalid model selected.", "Invalid model selected."
         return
@@ -159,6 +172,9 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "GLM-4.1V-9B-Thinking":
         processor = processor_s
         model = model_s
+    elif model_name == "kanana-1.5-v-3b":
+        processor = processor_f
+        model = model_f
     else:
         yield "Invalid model selected.", "Invalid model selected."
         return
@@ -273,7 +289,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
                     markdown_output = gr.Markdown(label="(Result.md)")
 
             model_choice = gr.Radio(
-                choices=["Camel-Doc-OCR-062825", "GLM-4.1V-9B-Thinking", "Megalodon-OCR-Sync-0713", "MonkeyOCR-pro-1.2B"],
+                choices=["Camel-Doc-OCR-062825", "GLM-4.1V-9B-Thinking", "Megalodon-OCR-Sync-0713", "MonkeyOCR-pro-1.2B", "kanana-1.5-v-3b"],
                 label="Select Model",
                 value="Camel-Doc-OCR-062825"
             )