Turn a story in Markdown into an animated reading film. The skill plans consistent characters and settings, creates one painting for each story beat, adds Fish Audio narration and timed captions, then renders and verifies the film.
Turns a story markdown file into a narrated animated film, a new reading format where the listener is read to and sees the story as it is told: beats from paragraphs, a bible so every character and place looks the same in every shot, one painting per beat in one style, Fish Audio narration in the story's voice with word timings, a HyperFrames film with camera moves, dissolves, scene breaks, a caption line and title cards, read back and verified. USE WHEN animated reading, story film, film this story, animate this story, illustrated audiobook, narrated story video, turn this story into a movie, storyboard this story, paint the shots, render the story film. NOT FOR the principal's own recordings, argument-first video essays, one-off motion graphics (Remotion), single illustrations (Art), writing the story itself (WriteStory).
Run score Not tested
See how to installStoryFilm turns a story written in Markdown into an animated film that combines narration and visuals. It divides the text into story beats, prepares reference material to keep characters and places consistent, and creates one painting for each beat. Fish Audio provides narration suited to the story. The film includes captions timed to the audio, camera movement, scene transitions, and title cards. The completed video is reviewed for issues such as mismatched images, captions, or audio. You can review the storyboard before production or ask the skill to proceed straight through.
Use it when you want readers to listen to a story while watching its scenes, or when you need a video whose images share a consistent look and character designs.
First-time setup requires Fish Audio and Google keys, as well as tools such as ffmpeg.
This overview was written by AI (OpenAI Codex) from the public SKILL.md. Check the original below for the exact behavior.
AI checked the SKILL.md for operations that need care and for setup requirements. The score comes only from actually running the skill (hands-on test).
Not counted in the score. Rated by AI (OpenAI Codex) as of Oct 28, 2026. Treat it as a guide and read the original before installing.
Review the instructions the agent will receive before you install.
A story goes in; a verified film of it comes out. One creative stage (the storyboard) and a deterministic tail (bible, shots, narration, render, verification, critique, delivery).
Out/verify.json green and complete, the Assay video eval set (video-eval) passed in the verify stage as Intents/Produce.md declares it, the review sheets looked at, and the ArtDirection critique passed on the exact bytes delivered.Run.ts --deliver, after a passing critique and a Proof receipt.| Intent | Trigger | File |
|---|---|---|
| Full | "film this story", "animated reading of X", "turn this story into a movie", anything that starts from a story file | Intents/Full.md |
| Storyboard | "storyboard this", "write the bible", "plan the shots", a redo of the creative stage after intake | Intents/Storyboard.md |
| Produce | "paint the shots", "render it", "re-render after an edit", "verify the film" | Intents/Produce.md |
| File | Read when |
|---|---|
References/StoryboardSchema.md | writing or editing Storyboard.json |
ISA.md | changing what the skill does: its ideal state, criteria and decisions |
Setup.md | first-time setup: Fish Audio and Google keys, ffmpeg and the other programs, optional Config.json |
Full is Intake → Storyboard → Produce. The storyboard is the review checkpoint when the principal wants one: it is cheap to change and forty paintings are not. Skip it by saying so ("straight through").
<run>/
run.json Beats.json Storyboard.json pronunciations.json (optional)
Bible/<id>.<hash>.png style plate + reference sheets
Shots/<beat>.<hash>.png one painting per beat · contact-sheet.png
Narration/<beat>.<hash>.{mp3,align.json,wav} narration.wav timeline.json
Out/composition/ the HyperFrames project Build.ts writes (index.html, plan.js, film.js, shots/, narration.wav)
Out/film.mp4 render.json render.log verify.json Critique.json frames/ caption-check/ review/{contact-sheet,motion-strip}.png
First-time setup (accounts, keys, files to fill in): Setup.md.
--voice, else the story's audio_voice frontmatter, else the DA's main voice from LIFEOS_CONFIG.toml. The value is a role (main, presenter, principal), fish:<reference id>, or the voice_id of a [da.voices.*] profile that has a fish_reference_id; any other id stops with an error naming where to add one. Narration is Fish Audio Drama 3 (FishTTS.ts) and takes no model flag; the DA's own voice reads at its configured fish_speed, any other voice at natural pace. Key: FISH_API_KEY in the env file; the principal role also needs FISH_PRINCIPAL_VOICE_ID.geminiImage lane in LIFEOS/TOOLS/models.ts on GOOGLE_API_KEY, 16:9 at 2K, reference images attached per shot.LIFEOS/USER/CUSTOMIZATIONS/SKILLS/StoryFilm/Config.json: runsRoot (default ~/StoryFilms; LIFEOS_STORYFILM_DIR overrides), fonts.serif / fonts.serifItalic (paths to licensed faces; path.ttc#<index> names one face of a collection), embedded into each composition as @font-face data. No font configured means the default serif, Source Serif 4 roman and italic (SIL Open Font License), installed by Setup.ts as the pinned @fontsource/source-serif-4 dependency.LIFEOS/USER/CUSTOMIZATIONS/SKILLS/StoryFilm/PREFERENCES.md: the principal's casting and register taste for the Storyboard intent. Config.json provingRun names the run slug the skill ISA's proving-run checks use (Setup.ts --proving-run prints it).USER/PRINCIPAL/PRONUNCIATIONS.json plus a per-run pronunciations.json, applied to spoken text only; the caption shows the story's words.Run.ts refuses a run dir under the config tree or the user-data repo (symlinks dereferenced first).Intake.ts runs again (2026-09-17: the narrator was re-cast in the story file after the first intake, and the first film went out in the old voice). Re-run Intake before any re-narration and diff the beat count.bun test <bare path> drops child-process output (Bun 1.3.6). Always bun test ./skills/StoryFilm/Tools.Example 1: a story file to a film
User: "Turn the story at stories/the-lighthouse.md into an animated reading"
→ Full: Intake writes Beats.json; the Storyboard intent writes the bible and one shot per beat; lint exits 0
→ Run.ts: bible painted, shots painted (4 at a time, cached by hash), narration synthesized, film rendered, verify.json green, review sheets written
→ ArtDirection critique of Out/film.mp4 passes into Out/Critique.json; Run.ts --from sheet --deliver lands the film and sheets in the downloads folder
Example 2: one painting is wrong
User: "The kitchen in beat 12 looks like a restaurant — fix it"
→ Edit that shot's prompt in Storyboard.json (and the place's bible desc if the sheet is the problem)
→ Run.ts --from lint --to shots repaints that one beat (everything else is cached); Run.ts --from build re-renders and re-verifies
Example 3: a character drifts
User: "Ana looks forty in half the shots"
→ Tighten her bible desc (age, face), Characters.ts --only ana repaints her sheet, Shots.ts --force repaints the shots that carry her ref (their hash changed with the sheet)
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Find and fetch vendor cybersecurity reports, then compare their findings on topics such as ransomware, breaches, cloud security, and AI. Analyses identify the reports consulted and cite each figure by vendor, year, and report, including where sources disagree.
Compare a component's own design needs with the goals of the larger system through three structured passes. The result identifies conflicts, recommends an appropriate scope, and assesses whether the proposed design fits the whole.
Find quotes that fit a text’s themes, with attribution and a short reason for each suggestion. Add sourced quotes, research thinkers, and organize a quote collection by theme and usage history to avoid repetition.
Create blog illustrations, technical diagrams, comparisons, comics, and icons through workflows tailored to each format. The skill guides composition, color, and model selection, then stages generated images for review before they are used in a project.
Assess the visual quality of images, films, slides, and site captures against a brief and craft criteria. Two vision judges score the work and rank defects, guiding bounded revision; delivery is gated on a passing critique, with human calibration labels recorded separately.
Find arXiv papers by topic, research category, or paper ID. Retrieve titles, authors, abstracts, and links, check new submissions, and add AlphaXiv's AI-generated overviews when available to help decide which papers to read.