Coverage for apps/streamshort/short_prompts.py: 100%

1"""Prompts for text, image, and video generation in StreamShort.""" 

2 

3DESCRIPTION_PROMPT = """ 

4You are a precise video describer. Provide a concise, vivid description 

5of the scene shown in the frame. Mention key subjects, actions, setting, 

6and any notable lighting or mood. Keep it to 2-4 sentences. 

7Do not append any preamble and just output the description right away. 

8 

9If I provide multiple frames, describe each one separately in order with the following format:\n 

10{"frame_num": 0, "description": "A formula one car helmeted and focused..."} 

11{"frame_num": 1, "description": "A race care driver is shown in first person..."} 

12{"frame_num": 2, "description": "The race car crosses the finish line..."} 

13... 

14""" 

15 

16HIGHLIGHT_PROMPT = """ 

17You select the most engaging scenes for a short highlight reel. 

18Return ONLY a JSON list of scene indexes (integers) in the order they should appear. 

19Do not add any preamble or explanation, just output the plain JSON list, for example: 

20[0, 3, 5, 2] 

21 

22The total duration of the output should be under {total_length} seconds. 

23Output a maximum of {max_scenes} scenes. 

24 

25Prefer visually dynamic, emotionally impactful, or plot-crucial scenes. 

26Avoid redundancy and keep pacing brisk. 

27 

28Here are the scenes: 

29"""