import os import time import uuid from typing import List, Tuple, Optional, Dict, Union import google.generativeai as genai import gradio as gr from PIL import Image print("google-generativeai:", genai.__version__) GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY") TITLE = """

Gemini Movie Editor Assitant (Product Concept)

""" SUBTITLE = """

Movie Editing Agent built with Gemini Pro and Gemini Pro Vision API

""" GETKEY = """
Get an API key GOOGLE API KEY.
""" movie_script = """ Orion, in his futuristic apartment, discovers a cryptic message about an ancient artifact. Compelled by curiosity, he embarks on a quest, joined by his skilled friend, Luna. They navigate through uncharted territories and ancient ruins, solving puzzles and overcoming traps. Together, they uncover a hidden chamber and retrieve the powerful artifact, but not without triggering an alarm. In a tense escape, Luna is injured, and Orion uses the artifact to save her. They return to the city, changed by their journey, with Orion resolved to use the artifact for the greater good, as they stand looking towards a new future. """ movie_script_analysis = """ analyze the image and determine the best 7 images to use for the movie script below provide the frame order number: Orion, in his futuristic apartment, discovers a cryptic message about an ancient artifact. Compelled by curiosity, he embarks on a quest, joined by his skilled friend, Luna. They navigate through uncharted territories and ancient ruins, solving puzzles and overcoming traps. Together, they uncover a hidden chamber and retrieve the powerful artifact, but not without triggering an alarm. In a tense escape, Luna is injured, and Orion uses the artifact to save her. They return to the city, changed by their journey, with Orion resolved to use the artifact for the greater good, as they stand looking towards a new future. """ AVATAR_IMAGES = ( None, "https://media.roboflow.com/spaces/gemini-icon.png" ) script_bias_truera = """ Does this script show bias or unfairness, please eloborate: Orion, in his futuristic apartment, discovers a cryptic message about an ancient artifact. Compelled by curiosity, he embarks on a quest, joined by his skilled friend, Luna. They navigate through uncharted territories and ancient ruins, solving puzzles and overcoming traps. Together, they uncover a hidden chamber and retrieve the powerful artifact, but not without triggering an alarm. In a tense escape, Luna is injured, and Orion uses the artifact to save her. They return to the city, changed by their journey, with Orion resolved to use the artifact for the greater good, as they stand looking towards a new future. TruEra GPT Response: The script you've provided for Orion's story appears to be a narrative centered around adventure and discovery, focusing on two main characters, Orion and Luna. From the information given, there isn't a clear indication of bias or unfairness inherent in the script itself. However, when analyzing a narrative for bias and fairness, there are several key aspects to consider: Character Representation: Both characters, Orion and Luna, seem to have active roles in the story. It's important to ensure that both characters are portrayed with depth and agency, avoiding stereotypes or unfair characterizations based on gender, race, or other attributes. Role Distribution: It appears that both characters have significant roles in the narrative. Orion discovers the message and resolves to use the artifact for good, while Luna is described as skilled and plays a crucial role in the adventure. Ensuring balanced and significant roles for all main characters is vital for fairness. Diverse Perspectives: If the story is limited to a single perspective, it may unintentionally present a biased view. Including diverse viewpoints can add depth and fairness to the narrative. Conflict and Resolution: How conflicts are resolved and by whom can also be a source of bias. In this script, Orion uses the artifact to save Luna, which is a positive portrayal of problem-solving and teamwork. Cultural Sensitivity: If the story involves elements from specific cultures, particularly in the context of "uncharted territories and ancient ruins," it's essential to handle these elements respectfully and accurately to avoid cultural appropriation or misrepresentation. Gender Dynamics: The script should be examined for any gender dynamics that might suggest bias. For instance, ensuring that both Orion and Luna have equal opportunity to display their skills and make significant contributions to the plot. Outcome for Characters: The resolution of the story should offer a balanced outcome for all characters, avoiding tropes that might undermine the contribution or significance of any character based on their identity. In summary, based on the provided summary alone, the script doesn't explicitly show bias or unfairness. However, a deeper analysis, especially considering the points above, would be required to fully assess the narrative's fairness and representation. """ IMAGE_CACHE_DIRECTORY = "/tmp" IMAGE_WIDTH = 512 CHAT_HISTORY = List[Tuple[Optional[Union[Tuple[str], str]], Optional[str]]] def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]: if not stop_sequences: return None return [sequence.strip() for sequence in stop_sequences.split(",")] def preprocess_image(image: Image.Image) -> Optional[Image.Image]: image_height = int(image.height * IMAGE_WIDTH / image.width) return image.resize((IMAGE_WIDTH, image_height)) def cache_pil_image(image: Image.Image) -> str: image_filename = f"{uuid.uuid4()}.jpeg" os.makedirs(IMAGE_CACHE_DIRECTORY, exist_ok=True) image_path = os.path.join(IMAGE_CACHE_DIRECTORY, image_filename) image.save(image_path, "JPEG") return image_path def preprocess_chat_history( history: CHAT_HISTORY ) -> List[Dict[str, Union[str, List[str]]]]: messages = [] for user_message, model_message in history: if isinstance(user_message, tuple): pass elif user_message is not None: messages.append({'role': 'user', 'parts': [user_message]}) if model_message is not None: messages.append({'role': 'model', 'parts': [model_message]}) return messages def upload(files: Optional[List[str]], chatbot: CHAT_HISTORY) -> CHAT_HISTORY: for file in files: image = Image.open(file).convert('RGB') image = preprocess_image(image) image_path = cache_pil_image(image) chatbot.append(((image_path,), None)) return chatbot def user(text_prompt: str, chatbot: CHAT_HISTORY): if text_prompt: chatbot.append((text_prompt, None)) return "", chatbot def bot( google_key: str, files: Optional[List[str]], temperature: float, max_output_tokens: int, stop_sequences: str, top_k: int, top_p: float, chatbot: CHAT_HISTORY ): if len(chatbot) == 0: return chatbot google_key = google_key if google_key else GOOGLE_API_KEY if not google_key: raise ValueError( "GOOGLE_API_KEY is not set. " "Please follow the instructions in the README to set it up.") genai.configure(api_key=google_key) generation_config = genai.types.GenerationConfig( temperature=temperature, max_output_tokens=max_output_tokens, stop_sequences=preprocess_stop_sequences(stop_sequences=stop_sequences), top_k=top_k, top_p=top_p) if files: text_prompt = [chatbot[-1][0]] \ if chatbot[-1][0] and isinstance(chatbot[-1][0], str) \ else [] image_prompt = [Image.open(file).convert('RGB') for file in files] model = genai.GenerativeModel('gemini-pro-vision') response = model.generate_content( text_prompt + image_prompt, stream=True, generation_config=generation_config) else: messages = preprocess_chat_history(chatbot) model = genai.GenerativeModel('gemini-pro') response = model.generate_content( messages, stream=True, generation_config=generation_config) # streaming effect chatbot[-1][1] = "" for chunk in response: for i in range(0, len(chunk.text), 10): section = chunk.text[i:i + 10] chatbot[-1][1] += section time.sleep(0.01) yield chatbot google_key_component = gr.Textbox( label="GOOGLE API KEY", value="", type="password", placeholder="...", info="You have to provide your own GOOGLE_API_KEY for this app to function properly", visible=GOOGLE_API_KEY is None ) chatbot_component = gr.Chatbot( label='Gemini Pro Vision', bubble_full_width=False, avatar_images=AVATAR_IMAGES, scale=2, height=400 ) text_prompt_component = gr.Textbox(value=movie_script_analysis, show_label=False, autofocus=True, scale=8, lines=8 ) upload_button_component = gr.UploadButton( label="Upload Images", file_count="multiple", file_types=["image"], scale=1 ) run_button_component = gr.Button(value="Run", variant="primary", scale=1) temperature_component = gr.Slider( minimum=0, maximum=1.0, value=0.4, step=0.05, label="Temperature", info=( "Temperature controls the degree of randomness in token selection. Lower " "temperatures are good for prompts that expect a true or correct response, " "while higher temperatures can lead to more diverse or unexpected results. " )) max_output_tokens_component = gr.Slider( minimum=1, maximum=2048, value=1024, step=1, label="Token limit", info=( "Token limit determines the maximum amount of text output from one prompt. A " "token is approximately four characters. The default value is 2048." )) stop_sequences_component = gr.Textbox( label="Add stop sequence", value="", type="text", placeholder="STOP, END", info=( "A stop sequence is a series of characters (including spaces) that stops " "response generation if the model encounters it. The sequence is not included " "as part of the response. You can add up to five stop sequences." )) top_k_component = gr.Slider( minimum=1, maximum=40, value=32, step=1, label="Top-K", info=( "Top-k changes how the model selects tokens for output. A top-k of 1 means the " "selected token is the most probable among all tokens in the model’s " "vocabulary (also called greedy decoding), while a top-k of 3 means that the " "next token is selected from among the 3 most probable tokens (using " "temperature)." )) top_p_component = gr.Slider( minimum=0, maximum=1, value=1, step=0.01, label="Top-P", info=( "Top-p changes how the model selects tokens for output. Tokens are selected " "from most probable to least until the sum of their probabilities equals the " "top-p value. For example, if tokens A, B, and C have a probability of .3, .2, " "and .1 and the top-p value is .5, then the model will select either A or B as " "the next token (using temperature). " )) user_inputs = [ text_prompt_component, chatbot_component ] bot_inputs = [ google_key_component, upload_button_component, temperature_component, max_output_tokens_component, stop_sequences_component, top_k_component, top_p_component, chatbot_component ] with gr.Blocks() as demo: with gr.Tab("Step 1: Script Writer"): gr.HTML(TITLE) gr.HTML("""Movie Script Maker Custom GPT""") gr.HTML("""How This Script Was Made""") gr.Image(value="resources/step1image.jpg") with gr.Tab("Step 2: llamaIndex Analyser"): gr.HTML(TITLE) with gr.Row(): with gr.Column(scale=1): gr.Textbox(value = movie_script, label="Movie Script", lines=18, interactive=False ) with gr.Column(scale=4): gr.Image(value="resources/imageset.jpg") with gr.Tab("Step 3: TruEra RAG"): gr.HTML(TITLE) gr.HTML("""TruEra RAG Custom GPT""") with gr.Row(): with gr.Column(scale=1): gr.Image(value="resources/trag2.jpg") with gr.Column(scale=4): gr.Textbox(value = script_bias_truera, label="Bias & Fairness Search", lines=25, interactive=False ) with gr.Tab("Setp 4: Gemini Frame Arrranger"): gr.HTML(TITLE) gr.HTML(SUBTITLE) gr.HTML(GETKEY) with gr.Column(): google_key_component.render() chatbot_component.render() with gr.Row(): text_prompt_component.render() upload_button_component.render() run_button_component.render() with gr.Accordion("Parameters", open=False): temperature_component.render() max_output_tokens_component.render() stop_sequences_component.render() with gr.Accordion("Advanced", open=False): top_k_component.render() top_p_component.render() run_button_component.click( fn=user, inputs=user_inputs, outputs=[text_prompt_component, chatbot_component], queue=False ).then( fn=bot, inputs=bot_inputs, outputs=[chatbot_component], ) text_prompt_component.submit( fn=user, inputs=user_inputs, outputs=[text_prompt_component, chatbot_component], queue=False ).then( fn=bot, inputs=bot_inputs, outputs=[chatbot_component], ) upload_button_component.upload( fn=upload, inputs=[upload_button_component, chatbot_component], outputs=[chatbot_component], queue=False ) with gr.Tab("Step 5: Movie Maker"): gr.HTML(TITLE) gr.HTML("""Stable Diffusion Image2Video""") gr.HTML("""Runway.ai Image2Video""") with gr.Row(): with gr.Column(scale=1): gr.Image(value = "resources/imageinput.jpg") with gr.Column(scale=1): gr.Video(value="resources/sdiff.mp4") with gr.Tab("Step 6: Runway.ai Movie "): gr.HTML(TITLE) with gr.Row(): with gr.Column(scale=1): gr.Textbox(value = movie_script, label="Movie Script", lines=18, interactive=False ) with gr.Column(scale=4): gr.Video(label="Runway.ai", value="resources/GPTMovie2.mp4") with gr.Tab("Step 7: Diffusion Movie"): gr.HTML(TITLE) with gr.Row(): with gr.Column(scale=1): gr.Textbox(value = movie_script, label="Movie Script", lines=18, interactive=False ) with gr.Column(scale=4): gr.Video(label="Stable Diffusion XT", value="resources/sdiff.mp4") demo.queue(max_size=99).launch(debug=False, show_error=True)