Update workflow to use Nemotron Prompt Generator and Ideogram 4 Image Generator

app.py

@@ -1,56 +1,74 @@
 import os
-import random
-
+import json
 import gradio as gr
+from gradio_client import Client
 
 
-def describe_product(image) -> str:
-    if image is None:
-        return "a premium product"
-    img_path = image.get("path") if isinstance(image, dict) else str(image)
-    img_url = image.get("url", "") if isinstance(image, dict) else str(image)
+def generate_prompt(concept: str) -> str:
+    """
+    Expands a simple concept into a detailed image prompt using the NVIDIA Nemotron model.
+    """
+    if not concept:
+        return "a ginger cat wearing a tiny wizard hat reading a spellbook"
     try:
-        from gradio_client import Client, handle_file
-        source = img_path if (img_path and os.path.exists(img_path)) else img_url
-        if not source:
-            return "a premium product"
-        client = Client("vikhyatk/moondream2", verbose=False)
+        client = Client("akhaliq/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4", verbose=False)
+        system_instruction = (
+            "You are an expert prompt engineer for text-to-image models. "
+            "Your task is to take a simple concept and expand it into a detailed, "
+            "vivid, and high-quality image prompt for Ideogram. "
+            "Describe the scene, lighting, materials, and aesthetic in detail. "
+            "Provide ONLY the final prompt text. Do not include any introductory or concluding text, "
+            "do not provide multiple options, and do not wrap the prompt in quotes."
+        )
+        messages = [
+            {"role": "system", "content": system_instruction},
+            {"role": "user", "content": f"Concept: {concept}"}
+        ]
         result = client.predict(
-            handle_file(source),
-            "Describe this product briefly and concisely. What is it?",
-            api_name="/answer_question",
+            messages_json=json.dumps(messages),
+            temperature=0.7,
+            max_tokens=256,
+            custom_token=None,
+            api_name="/chat"
         )
-        return str(result).strip() if result else "a premium product"
-    except Exception:
-        return "a premium product"
+        clean_result = str(result).strip()
+        if clean_result.startswith('"') and clean_result.endswith('"'):
+            clean_result = clean_result[1:-1]
+        elif clean_result.startswith("'") and clean_result.endswith("'"):
+            clean_result = clean_result[1:-1]
+        return clean_result
+    except Exception as e:
+        print(f"Error calling Nemotron space: {e}")
+        return f"A detailed, high-quality, professional commercial product photograph of {concept}"
 
 
-def craft_marketing_prompt(caption: str) -> str:
-    if not caption:
-        caption = "a premium product"
-    caption = caption.strip().rstrip(".")
-    styles = [
-        (
-            f"Professional product advertisement photograph of {caption}, "
-            "studio lighting, clean white background, commercial photography, "
-            "ultra-sharp, 8K quality"
-        ),
-        (
-            f"Cinematic product shot of {caption}, dramatic lighting, "
-            "aspirational lifestyle context, premium brand aesthetic, "
-            "shot on Hasselblad, magazine cover quality"
-        ),
-        (
-            f"Bold marketing campaign visual of {caption}, vibrant colors, "
-            "dynamic composition, modern editorial style, "
-            "award-winning commercial photography"
-        ),
-    ]
-    return random.choice(styles)
+def generate_image(prompt: str) -> str:
+    """
+    Generates an image from a prompt using the Ideogram 4 model.
+    """
+    if not prompt:
+        prompt = "a ginger cat wearing a tiny wizard hat reading a spellbook"
+    try:
+        client = Client("ideogram-ai/ideogram4", verbose=False)
+        result = client.predict(
+            prompt=prompt,
+            mode="Default · 20 steps",
+            upsampler="Ideogram (remote)",
+            width=1024,
+            height=1024,
+            seed=0,
+            randomize_seed=True,
+            api_name="/generate"
+        )
+        if isinstance(result, (list, tuple)) and len(result) > 0:
+            return result[0]
+        return result
+    except Exception as e:
+        print(f"Error calling Ideogram 4 space: {e}")
+        raise e
 
 
-demo = gr.Workflow(bind=[describe_product, craft_marketing_prompt])
+demo = gr.Workflow(bind=[generate_prompt, generate_image])
 
 if __name__ == "__main__":
     demo.launch()
-

workflow.json

@@ -1 +1,150 @@
-{"schema_version":"2","name":"Marketing Image Creator","runtime":{"default":"client"},"references":[{"label":"Text","inputs":[{"id":"in","label":"Text","type":"text"}],"outputs":[{"id":"out","label":"Text","type":"text"}],"width":220,"height":163,"asset_type":"text","role":"reference","id":"1539a1e9-4906-4008-b345-e6d05bb69b22","x":740,"y":80,"data":{"in":null}},{"label":"Image","inputs":[{"id":"in","label":"Image","type":"image"}],"outputs":[{"id":"out","label":"Image","type":"image"}],"width":220,"height":124,"asset_type":"image","role":"reference","id":"13e92b22-7173-416c-81c6-1ca097a2a299","x":80,"y":80,"data":{}}],"operators":[{"id":"n_flux","label":"Generate Image","inputs":[{"id":"in","label":"Prompt","type":"text","required":true}],"outputs":[{"id":"out","label":"Image","type":"image"}],"data":{"out":97500097},"x":1040,"y":80,"width":220,"height":124,"role":"operator","kind":"space","source":"hf://spaces/multimodalart/FLUX.1-merged","space_id":"multimodalart/FLUX.1-merged","runtime":"client","endpoints":[{"name":"/infer","inputs":[{"id":"in_0","label":"Prompt","type":"text","required":true},{"id":"in_1","label":"Seed","type":"number","required":false},{"id":"in_2","label":"Randomize seed","type":"boolean","required":false},{"id":"in_3","label":"Width","type":"number","required":false},{"id":"in_4","label":"Height","type":"number","required":false},{"id":"in_5","label":"Guidance Scale","type":"number","required":false},{"id":"in_6","label":"Number of inference steps","type":"number","required":false}],"outputs":[{"id":"out_0","label":"Result","type":"image","output_index":0},{"id":"out_1","label":"Seed","type":"number","output_index":1}]}]},{"label":"Image To Prompt","inputs":[{"id":"in_0","label":"Input Image","type":"image","required":true}],"outputs":[{"id":"out_0","label":"Output Prompt","type":"text","output_index":0}],"width":280,"height":124,"kind":"space","space_id":"ovi054/image-to-prompt","endpoint":"/predict","endpoints":[{"name":"/predict","inputs":[{"id":"in_0","label":"Input Image","type":"image","required":true}],"outputs":[{"id":"out_0","label":"Output Prompt","type":"text","output_index":0}]}],"pipeline_tag":"Image Captioning","role":"operator","id":"897592dc-17b5-473b-9e89-4ebdae4073d6","x":380,"y":80,"data":{}}],"subjects":[{"id":"n_output","label":"Marketing Image","inputs":[{"id":"in","label":"Image","type":"image"}],"outputs":[{"id":"out","label":"Image","type":"image"}],"data":{"in":null},"x":1340,"y":80,"width":220,"height":107,"role":"subject","asset_type":"image"}],"edges":[{"id":"e4","from_node_id":"n_flux","from_port_id":"out","to_node_id":"n_output","to_port_id":"in","type":"image"},{"from_node_id":"1539a1e9-4906-4008-b345-e6d05bb69b22","from_port_id":"out","to_node_id":"n_flux","to_port_id":"in","type":"text","id":"b44e83fe-23b1-4dcc-83d3-55b3481721bc"},{"from_node_id":"13e92b22-7173-416c-81c6-1ca097a2a299","from_port_id":"out","to_node_id":"897592dc-17b5-473b-9e89-4ebdae4073d6","to_port_id":"in_0","type":"image","id":"22796709-ff7d-4250-92c0-2d541f9e6a99"},{"from_node_id":"897592dc-17b5-473b-9e89-4ebdae4073d6","from_port_id":"out_0","to_node_id":"1539a1e9-4906-4008-b345-e6d05bb69b22","to_port_id":"in","type":"text","id":"96dab823-0f2c-4356-98c4-aa82feca5e0a"}],"view":{"default":"canvas"}}
+{
+  "schema_version": "2",
+  "name": "Ideogram 4 Workflow",
+  "runtime": {
+    "default": "client"
+  },
+  "references": [
+    {
+      "id": "ref_concept",
+      "label": "Text",
+      "inputs": [
+        {
+          "id": "in",
+          "label": "Text",
+          "type": "text"
+        }
+      ],
+      "outputs": [
+        {
+          "id": "out",
+          "label": "Text",
+          "type": "text"
+        }
+      ],
+      "width": 220,
+      "height": 163,
+      "asset_type": "text",
+      "role": "reference",
+      "data": {
+        "in": "a ginger cat wearing a tiny wizard hat reading a spellbook"
+      },
+      "x": 80,
+      "y": 80
+    }
+  ],
+  "operators": [
+    {
+      "id": "op_generate_prompt",
+      "label": "Nemotron Prompt Generator",
+      "role": "operator",
+      "kind": "fn",
+      "fn": "generate_prompt",
+      "inputs": [
+        {
+          "id": "in_0",
+          "label": "concept",
+          "type": "text",
+          "required": true
+        }
+      ],
+      "outputs": [
+        {
+          "id": "out_0",
+          "label": "output",
+          "type": "text"
+        }
+      ],
+      "data": {},
+      "x": 380,
+      "y": 80,
+      "width": 280,
+      "height": 124
+    },
+    {
+      "id": "op_generate_image",
+      "label": "Ideogram 4 Image Generator",
+      "role": "operator",
+      "kind": "fn",
+      "fn": "generate_image",
+      "inputs": [
+        {
+          "id": "in_0",
+          "label": "prompt",
+          "type": "text",
+          "required": true
+        }
+      ],
+      "outputs": [
+        {
+          "id": "out_0",
+          "label": "output",
+          "type": "image"
+        }
+      ],
+      "data": {},
+      "x": 740,
+      "y": 80,
+      "width": 280,
+      "height": 124
+    }
+  ],
+  "subjects": [
+    {
+      "id": "subj_output_image",
+      "label": "Generated Image",
+      "role": "subject",
+      "asset_type": "image",
+      "inputs": [
+        {
+          "id": "in",
+          "label": "Image",
+          "type": "image"
+        }
+      ],
+      "outputs": [
+        {
+          "id": "out",
+          "label": "Image",
+          "type": "image"
+        }
+      ],
+      "data": {
+        "in": null
+      },
+      "x": 1100,
+      "y": 80,
+      "width": 220,
+      "height": 107
+    }
+  ],
+  "edges": [
+    {
+      "id": "edge_1",
+      "from_node_id": "ref_concept",
+      "from_port_id": "out",
+      "to_node_id": "op_generate_prompt",
+      "to_port_id": "in_0",
+      "type": "text"
+    },
+    {
+      "id": "edge_2",
+      "from_node_id": "op_generate_prompt",
+      "from_port_id": "out_0",
+      "to_node_id": "op_generate_image",
+      "to_port_id": "in_0",
+      "type": "text"
+    },
+    {
+      "id": "edge_3",
+      "from_node_id": "op_generate_image",
+      "from_port_id": "out_0",
+      "to_node_id": "subj_output_image",
+      "to_port_id": "in",
+      "type": "image"
+    }
+  ],
+  "view": {
+    "default": "canvas"
+  }
+}