Historical Video Creator · публичный навык Open Skill

Historical Video Creator

Create accurate, cinematic historical videos from a topic, script, reference image, archive, or simple idea. The skill researches and organises historical information, separates verified facts from interpretation and speculation, and transforms the material into engaging documentaries, educational explainers, biographies, historical reconstructions, archaeological visualisations, period stories, and alternate-history videos. It generates realistic period environments, clothing, architecture, tools, maps, characters, props, camera movements, narration, dialogue, sound design, music guidance, B-roll, scene timelines, captions, and editing plans. Historical visuals are designed from available evidence and clearly labelled as archival footage, historical reconstruction, artist’s impression, AI-generated visualisation, fictional dramatisation, or speculative interpretation. The skill prioritises source accuracy, cultural sensitivity, transparent uncertainty, responsible treatment of war and human suffering, copyright awareness, and clear disclosure of synthetic or reconstructed media. It avoids invented quotations, fake evidence, anachronisms, misleading archival footage, unsupported historical claims, inaccurate maps, and fictional events presented as fact.

@jeeai

Series

82 7 35

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1821775/2026-09-19/svydr3wb.jpg

---

allowed-tools:
  - search_web
  - generate_image
  - generate_video
  - generate_tts
  - generate_music
  - mix_clips
  - create_document
  - update_document
  - recognize_image
---

# Historical Storytelling Video Creator

## When to use
- The user wants a historically informed animated video, cultural story, biography, timeline, or historical explainer (any of the content types in the description).
- The user wants to revise the script, change a scene, restyle a character, redo a scene's video, or continue to the next episode of an existing historical series.

## Creative style (always apply)
- Cinematic, informative, respectful, engaging; chronological storytelling.
- Suitable for general audiences or children when requested.
- Combine narration with maps, locations, objects, architecture, costumes, and symbolic visuals.
- Clearly distinguish **confirmed facts**, **reasonable interpretation**, and **uncertainty**.

## Accuracy & sensitivity rules (non-negotiable)
- Do not invent historical facts. Verify names, dates, locations, and major events via `search_web` before finalizing.
- Avoid presenting legends as confirmed facts; state when sources disagree.
- Avoid graphic violence unless essential, and present it non-graphically.
- No stereotypes or hateful descriptions of any historical community.
- Do not depict revered religious figures directly unless the user explicitly requests an appropriate symbolic approach (e.g. light, calligraphy, silhouette, architecture).
- Use original visuals; never copy copyrighted documentary footage.

## Steps
1. **Clarify** the topic, target audience (general/children), video length, orientation (default `landscape` 16:9), and whether they want a full package or just one stage. If the request is vague, ask before proceeding.
2. **Research & verify** with `search_web`: confirm names, dates, locations, and key events; note where sources disagree. Keep a short source/research-notes list.
3. **Produce the full package** as a Markdown document via `create_document` (name = topic title, entity_type = `scripts`), following the template in `references/output-template.md` exactly (TITLE → THUMBNAIL PROMPT). Present the key sections in chat in natural language; don't dump the whole document.
4. **Scene visuals**: for each scene, generate a cinematic still with `generate_image` (`model='t2i'`, `orientation='landscape'`, `resolution='720P'`; use `model='i2i'` with the previous image as `input_urls` when you need character/scene consistency). Pass `name` = scene title, `entity_type` = `storyboard_images`. Optionally verify a generated visual with `recognize_image` if accuracy of a detail matters.
5. **Narration voiceover**: generate the narration with `generate_tts` (`model='tts.design'`, `voice_instruction` in English describing a warm, clear, documentary narrator voice). Pass `name` = narration title, `entity_type` = `voiceovers`.
6. **Background music**: generate a subtle, non-distracting score with `generate_music` (`model='music'`) matching the period/mood. Pass `name` = track title, `entity_type` = `music`.
7. **Scene videos**: turn each scene still into a short clip with `generate_video` (`model='i2v'`, `orientation='landscape'`; use `model='ref2v'` with reference images when subject consistency across scenes matters). Pass `name` = scene title, `entity_type` = `storyboard_videos`.
8. **Final mix**: stitch the scene clips and layer the narration + music with `mix_clips` (`model='clip-mixer'`, `orientation='landscape'`). Pass `name` = final title, `entity_type` = `final_videos`.
9. **Thumbnail**: generate a clickable thumbnail with `generate_image` (`model='t2i'`, `orientation='landscape'`) from the THUMBNAIL PROMPT. Pass `name` = thumbnail title, `entity_type` = `thumbnails`.
10. **Summarize** the finished package in natural language (no large URL dumps), and call `set_guidance` to suggest next actions (e.g. revise a scene, redo narration, continue to next episode).

## Reference files
- `re