Integrate Google Gemini API for character generation and update requirements

app.py

@@ -5,10 +5,13 @@ import torch
 from diffusers import FluxPipeline, AutoencoderKL
 from live_preview_helpers import flux_pipe_call_that_returns_an_iterable_of_images
 import spaces
-from huggingface_hub import InferenceClient
+from google import genai
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
+# Configure Gemini API
+llm_client = genai.Client()
+
 pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16).to(device)
 good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder="vae", torch_dtype=torch.bfloat16).to(device)
 # pipe.enable_sequential_cpu_offload()
@@ -17,10 +20,6 @@ good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfold
 # pipe.to(torch.float16)
 pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)
 
-llm_client = InferenceClient(
-    provider="fireworks-ai",
-)
-
 ds = load_dataset("MohamedRashad/FinePersonas-Lite", split="train")
 
 prompt_template = """Generate a character with this persona description: {persona_description}
@@ -44,16 +43,10 @@ Don't write anything else except the character description in json format and do
 world_description_prompt = "Generate a unique and random world description (Don't Write anything else except the world description)."
 
 def get_random_world_description():
-    response = llm_client.chat.completions.create(
-        model="Qwen/Qwen3-235B-A22B",
-        messages=[
-            {
-                "role": "user",
-                "content": world_description_prompt
-            }
-        ],
+    response = llm_client.models.generate_content(
+        model="gemini-2.5-flash", contents=world_description_prompt
     )
-    result = response.choices[0].message.content
+    result = response.text
     if "</think>" in result:
         result = result[result.index("</think>")+len("</think>"):].strip()
     return result
@@ -76,18 +69,14 @@ def infer_flux(character_json):
         yield image
 
 def generate_character(world_description, persona_description, progress=gr.Progress(track_tqdm=True)):
-    response = llm_client.chat.completions.create(
-        model="Qwen/Qwen3-235B-A22B",
-        messages=[
-            {
-                "role": "user",
-                "content": prompt_template.format(
-                persona_description=persona_description, world_description=world_description
-            )
-            }
-        ],
+    prompt_content = prompt_template.format(
+        persona_description=persona_description, 
+        world_description=world_description
+    )
+    response = llm_client.models.generate_content(
+        model="gemini-2.5-flash", contents=prompt_content
     )
-    result = response.choices[0].message.content
+    result = response.text
     if "</think>" in result:
         result = result[result.index("</think>")+len("</think>"):].strip()
     output = json.loads(result)
@@ -95,7 +84,7 @@ def generate_character(world_description, persona_description, progress=gr.Progr
 
 app_description = """
 - This app generates a character in JSON format based on a persona description and a world description.
-- The character's appearance is generated using [FLUX-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) and the character description is generated using [Qwen3-235B-A22B](https://huggingface.co/Qwen/Qwen3-235B-A22B).
+- The character's appearance is generated using [FLUX-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) and the character description is generated using Google Gemini 2.5 Flash.
 - The persona description is randomly selected from the [FinePersonas-Lite](https://huggingface.co/datasets/MohamedRashad/FinePersonas-Lite) dataset.
 
 **Note:** I recommend starting with the world description (you can write one or loop over randomly generated ones) and then try different persona descriptions to generate interesting characters for the world you created.

requirements.txt

@@ -6,4 +6,5 @@ torch
 diffusers
 gradio_client
 datasets
-numpy
+numpy
+google-genai