Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

You are a developer assistant where you only provide the code…

  1. User

    Assistant: User: System: You are a developer assistant where you only provide the code for a question. No explanation required. Write a simple json sample. Given the transcript segments, can you generate a list of highlights with start and end times for the video using multiple segments? Please meet the following constraints:

    - The highlights should be a direct part of the video and should not be out of context.
    - The highlights should be interesting and clippable, providing value to the viewer.
    - The highlights should be 20 to 60 seconds MAXIMUM.
    - The highlights should include more than one segment to provide context and continuity.
    - The highlights should not cut off in the middle of a sentence or idea.
    - The user provided highlight phrases should be used to generate the highlights
    - The highlights should be based on the relevance of the segments to the highlight phrases
    - The highlights should be scored out of 100 based on the relevance of the segments to the highlight phrases.
    
    Respond with the following JSON schema with valid JSON syntax for highlights:
    
    {'$defs': {'Highlight': {'properties': {'title': {'description': 'Title of the Video highlight', 'title': 'Title', 'type': 'string'}, 'start_time': {'description': 'Start time of the video highlight', 'title': 'Start Time', 'type': 'number'}, 'end_time': {'description': 'End time of the video highlight', 'title': 'End Time', 'type': 'number'}, 'score': {'description': 'Score of the video highlight', 'title': 'Score', 'type': 'number'}}, 'required': ['title', 'start_time', 'end_time', 'score'], 'title': 'Highlight', 'type': 'object'}}, 'properties': {'chapters': {'description': 'List of Highlights', 'items': {'$ref': '#/$defs/Highlight'}, 'title': 'Chapters', 'type': 'array'}}, 'required': ['chapters'], 'title': 'HighlightSchema', 'type': 'object'}
    
    Each highlight should have the following fields:
    - title: title of the highlight
    - start_time: start time of the highlight as a float in seconds
    - end_time: end time of the highlight as a float in seconds
    - score: score of the highlight as a float out of 100
    

    User: Key Phrases: Funny

    going to be like among us are just random dudes right now. And you think that [568.26 - 573.10]: there's orders already. Everybody thinks it's them. Trust me. It's not me. It's not me. I got the [573.10 - 579.58]: West Point. I found out right away. It wasn't me. He's soft. There's dudes that are built for the [579.58 - 585.24]: mud in the blood. But you think that they're just working like a union job. But right now, [585.24 - 589.24]: where do you think most of them are enlisted? No wall is a teacher. You never know. That's what I'm [589.24 - 596.76]: saying. Like when real shit hits the fan. Not these back rules. Those freaks can do it. Assistant: Assistant:

  2. Assistant
    {
      "chapters": [
        {
          "title": "Funny Conversation on Roles",
          "start_time": 568.26,
          "end_time": 596.76,
          "score": 85.0
        }
      ]
    }
    

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.