from functools import lru_cache import gradio as gr from gradio_client import Client MODEL_SPACE = "black-forest-labs/FLUX.1-schnell" DEFAULT_PROMPT = ( "A highly realistic photograph of a peaceful working farm in the Canadian " "countryside at golden hour, a red wooden barn, green crop fields, a tractor, " "a few cows grazing behind a wooden fence, warm natural light, true-to-life " "colors, detailed textures, photorealistic, no text, no logo" ) @lru_cache(maxsize=1) def flux_client() -> Client: """Reuse one connection to the official FLUX.1-schnell Space.""" # ZeroGPU Spaces can have a queue. A generous timeout prevents the request # from failing while it waits for an available GPU. return Client(MODEL_SPACE, httpx_kwargs={"timeout": 180.0}) def generate_image( prompt: str, seed: int, randomize_seed: bool, width: int, height: int, steps: int, ): prompt = prompt.strip() if not prompt: raise gr.Error("Veuillez écrire une description de l’image.") try: result = flux_client().predict( prompt=prompt, seed=int(seed), randomize_seed=bool(randomize_seed), width=int(width), height=int(height), num_inference_steps=int(steps), api_name="/infer", ) except Exception as exc: raise gr.Error( "La génération FLUX.1 est momentanément indisponible. " "Veuillez réessayer dans quelques instants." ) from exc if isinstance(result, (tuple, list)): image = result[0] used_seed = result[1] if len(result) > 1 else seed else: image = result used_seed = seed if isinstance(image, dict): image = image.get("path") or image.get("url") return image, int(used_seed) with gr.Blocks(title="Générateur réaliste FLUX.1") as demo: gr.Markdown( """ # Générateur d’images réalistes FLUX.1 Décrivez l’image souhaitée, ajustez les paramètres si nécessaire, puis cliquez sur **Générer l’image**. """ ) with gr.Row(): with gr.Column(scale=2): prompt_input = gr.Textbox( label="Description de l’image", value=DEFAULT_PROMPT, lines=7, placeholder="Exemple : une ferme réaliste au coucher du soleil…", ) with gr.Accordion("Paramètres avancés", open=False): seed_input = gr.Slider( minimum=0, maximum=2_147_483_647, value=18_374_629, step=1, label="Graine", ) randomize_seed_input = gr.Checkbox( value=True, label="Utiliser une nouvelle graine à chaque génération", ) with gr.Row(): width_input = gr.Slider( minimum=256, maximum=1536, value=1024, step=32, label="Largeur", ) height_input = gr.Slider( minimum=256, maximum=1536, value=768, step=32, label="Hauteur", ) steps_input = gr.Slider( minimum=1, maximum=8, value=4, step=1, label="Nombre d’étapes", ) generate_button = gr.Button("Générer l’image", variant="primary") with gr.Column(scale=3): image_output = gr.Image(label="Image générée", type="filepath") seed_output = gr.Number(label="Graine utilisée", precision=0) generate_button.click( fn=generate_image, inputs=[ prompt_input, seed_input, randomize_seed_input, width_input, height_input, steps_input, ], outputs=[image_output, seed_output], api_name="generate", ) prompt_input.submit( fn=generate_image, inputs=[ prompt_input, seed_input, randomize_seed_input, width_input, height_input, steps_input, ], outputs=[image_output, seed_output], ) if __name__ == "__main__": demo.queue(default_concurrency_limit=2).launch() app.py from functools import lru_cache import gradio as gr from gradio_client import Client MODEL_SPACE = "black-forest-labs/FLUX.1-schnell" DEFAULT_PROMPT = ( "A highly realistic photograph of a peaceful working farm in the Canadian " "countryside at golden hour, a red wooden barn, green crop fields, a tractor, " "a few cows grazing behind a wooden fence, warm natural light, true-to-life " "colors, detailed textures, photorealistic, no text, no logo" ) @lru_cache(maxsize=1) def flux_client() -> Client: """Reuse one connection to the official FLUX.1-schnell Space.""" # ZeroGPU Spaces can have a queue. A generous timeout prevents the request # from failing while it waits for an available GPU. return Client(MODEL_SPACE, httpx_kwargs={"timeout": 180.0}) def generate_image( prompt: str, seed: int, randomize_seed: bool, width: int, height: int, steps: int, ): prompt = prompt.strip() if not prompt: raise gr.Error("Veuillez écrire une description de l’image.") try: result = flux_client().predict( prompt=prompt, seed=int(seed), randomize_seed=bool(randomize_seed), width=int(width), height=int(height), num_inference_steps=int(steps), api_name="/infer", ) except Exception as exc: raise gr.Error( "La génération FLUX.1 est momentanément indisponible. " "Veuillez réessayer dans quelques instants." ) from exc if isinstance(result, (tuple, list)): image = result[0] used_seed = result[1] if len(result) > 1 else seed else: image = result used_seed = seed if isinstance(image, dict): image = image.get("path") or image.get("url") return image, int(used_seed) with gr.Blocks(title="Générateur réaliste FLUX.1") as demo: gr.Markdown( """ # Générateur d’images réalistes FLUX.1 Décrivez l’image souhaitée, ajustez les paramètres si nécessaire, puis cliquez sur **Générer l’image**. """ ) with gr.Row(): with gr.Column(scale=2): prompt_input = gr.Textbox( label="Description de l’image", value=DEFAULT_PROMPT, lines=7, placeholder="Exemple : une ferme réaliste au coucher du soleil…", ) with gr.Accordion("Paramètres avancés", open=False): seed_input = gr.Slider( minimum=0, maximum=2_147_483_647, value=18_374_629, step=1, label="Graine", ) randomize_seed_input = gr.Checkbox( value=True, label="Utiliser une nouvelle graine à chaque génération", ) with gr.Row(): width_input = gr.Slider( minimum=256, maximum=1536, value=1024, step=32, label="Largeur", ) height_input = gr.Slider( minimum=256, maximum=1536, value=768, step=32, label="Hauteur", ) steps_input = gr.Slider( minimum=1, maximum=8, value=4, step=1, label="Nombre d’étapes", ) generate_button = gr.Button("Générer l’image", variant="primary") with gr.Column(scale=3): image_output = gr.Image(label="Image générée", type="filepath") seed_output = gr.Number(label="Graine utilisée", precision=0) generate_button.click( fn=generate_image, inputs=[ prompt_input, seed_input, randomize_seed_input, width_input, height_input, steps_input, ], outputs=[image_output, seed_output], api_name="generate", ) prompt_input.submit( fn=generate_image, inputs=[ prompt_input, seed_input, randomize_seed_input, width_input, height_input, steps_input, ], outputs=[image_output, seed_output], ) if __name__ == "__main__": demo.queue(default_concurrency_limit=2).launch()