Dataset Viewer
The dataset viewer is not available for this dataset.
Cannot get the config names for the dataset.
Error code:   ConfigNamesError
Exception:    ValueError
Message:      Feature type 'Structural' not found. Available feature types: ['Value', 'ClassLabel', 'Translation', 'TranslationVariableLanguages', 'LargeList', 'List', 'Array2D', 'Array3D', 'Array4D', 'Array5D', 'Audio', 'Image', 'Mesh', 'Video', 'Pdf', 'Nifti', 'BioSequence', 'BioStructure', 'Json']
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/dataset/config_names.py", line 67, in compute_config_names_response
                  config_names = get_dataset_config_names(
                      path=dataset,
                      token=hf_token,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 161, in get_dataset_config_names
                  dataset_module = dataset_module_factory(
                      path,
                  ...<4 lines>...
                      **download_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1182, in dataset_module_factory
                  raise e1 from None
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1157, in dataset_module_factory
                  ).get_module()
                    ~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 614, in get_module
                  dataset_infos = DatasetInfosDict.from_dataset_card_data(dataset_card_data)
                File "/usr/local/lib/python3.14/site-packages/datasets/info.py", line 399, in from_dataset_card_data
                  dataset_info = DatasetInfo._from_yaml_dict(dataset_card_data["dataset_info"])
                File "/usr/local/lib/python3.14/site-packages/datasets/info.py", line 317, in _from_yaml_dict
                  yaml_data["features"] = Features._from_yaml_list(yaml_data["features"])
                                          ~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2230, in _from_yaml_list
                  return cls.from_dict(from_yaml_inner(yaml_data))
                         ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2075, in from_dict
                  obj = generate_from_dict(dic)
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1648, in generate_from_dict
                  return {key: generate_from_dict(value) for key, value in obj.items()}
                               ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1661, in generate_from_dict
                  return List(generate_from_dict(feature), **obj)
                              ~~~~~~~~~~~~~~~~~~^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1654, in generate_from_dict
                  raise ValueError(f"Feature type '{_type}' not found. Available feature types: {list(_FEATURE_TYPES.keys())}")
              ValueError: Feature type 'Structural' not found. Available feature types: ['Value', 'ClassLabel', 'Translation', 'TranslationVariableLanguages', 'LargeList', 'List', 'Array2D', 'Array3D', 'Array4D', 'Array5D', 'Audio', 'Image', 'Mesh', 'Video', 'Pdf', 'Nifti', 'BioSequence', 'BioStructure', 'Json']

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:The task_categories "conversational" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

FlaneAI/Opus-5.5-YouTube-Video-Creation

A synthetic SFT dataset generated by Claude Opus 5.5, teaching models how to plan, script, and produce internet-style YouTube videos β€” from idea to final render β€” using code, tools, and step-by-step reasoning.

This dataset is NOT a dump of Claude's raw internal chain-of-thought. It contains all tool calls, tool outputs, Python animation code, and recreated reasoning traces that describe what the model did, not the hidden native reasoning.

What is this dataset?

This dataset captures Claude Opus 5.5 performing the complete workflow of creating internet-native YouTube videos β€” not cinematic movies. That means:

  • Internet-style content: YouTube videos, shorts, tutorials, commentary, explainers, vlogs
  • Python-driven creation: animations, motion graphics, visualizations, and renders written as executable code
  • Tool usage (OpenAI-compatible schema): realistic tool calls and their responses
  • Recreated reasoning: human-readable reasoning traces that were reconstructed to explain the actions taken β€” not the original hidden CoT

What's inside?

Each example is a conversation in OpenAI-style chat message format:

Role Content
system Instructions: create a YouTube-style video for the given prompt
assistant Recreated reasoning trace + natural language + tool calls
tool Tool call results (outputs, errors, files, render info)

Structure

{
  "messages": [
    {
      "role": "system",
      "content": "You are an AI video creation agent. Create an internet-style YouTube video..."
    },
    {
      "role": "assistant",
      "content": "Reasoning (recreated): First, I need to understand the target style...",
      "tool_calls": [
        {
          "id": "call_01",
          "type": "function",
          "function": {
            "name": "run_python_code",
            "arguments": "{\"code\": \"import matplotlib...\"}"
          }
        }
      ]
    },
    {
      "role": "tool",
      "tool_call_id": "call_01",
      "content": "<execution output>"
    },
    {
      "role": "assistant",
      "content": "Final video created: intro β†’ main content β†’ outro..."
    }
  ]
}

Tool calls covered

  • run_python_code β€” Matplotlib, Manim, MoviePy, Pillow animations & rendering
  • file_operations β€” writing, reading, organizing project files
  • video/render tools β€” assembling scenes, adding audio, exporting final video
  • web/search tools β€” research for video content, trends, facts

About the "reasoning traces"

A word on transparency:

  • These are NOT Claude Opus 5.5's original hidden chain-of-thought
  • They are recreated traces: reconstructed narratives explaining each action, tool choice, and code decision
  • This makes the data safe, clean, and fully documented β€” ideal for SFT

Use cases

  • SFT / fine-tuning of LLMs on tool-calling + code-generation behavior
  • Training video-creation agents (text β†’ YouTube video)
  • Teaching models to combine reasoning + Python + tool orchestration
  • Building AI content-creation assistants

πŸ“ License

apache-2.0 β€” free to use, modify, and distribute with attribution.

πŸ™ Acknowledgments

  • Generated with Claude Opus 5.5 (Anthropic)
  • Processed and curated by FlaneAI

Made with ❀️ by FlaneAI

Downloads last month
18