Modo
Agendar llamada

hyperframes: crear videos con HTML y renderizarlos a MP4

Hace que Claude arme videos como páginas HTML animadas (lanzamientos, explicativos, gráficos, clips verticales) y los convierta en un MP4 que sale igual cada vez.

hyperframes: crear videos con HTML y renderizarlos a MP4

Respuesta rápida

Qué es
hyperframes es una skill de HeyGen (licencia Apache 2.0) que le enseña a Claude a crear videos escribiendo HTML animado y a renderizarlos a MP4 cuadro por cuadro.
Qué le das
Qué video quieres, para qué red y el material de tu marca (logo, capturas, tu web, guion o voz)
Qué te devuelve
Un proyecto de video editable, una vista previa para revisar y el MP4 final
Autor
HeyGen
Licencia
Apache 2.0
En GitHub
84,8 mil estrellas · actualizada el
Revisada por Claura

Instálala en una línea para todos tus proyectos, en Claude Code, Cursor, Codex y más:

npx skills add heygen-com/hyperframes --skill hyperframes -g

Funciona en

Claude CodeEn ~/.claude/skills/
CursorEn ~/.agents/skills/
CodexEn ~/.agents/skills/
Gemini CLIEn ~/.agents/skills/

Pedidos para probarla

Con la skill instalada, copia uno y pégaselo a tu agente tal cual, o cambia los datos por los tuyos.

  • Haz un video de lanzamiento de 30 segundos de la nueva agenda online usando las pantallas de nuestra web, y sácalo también en vertical para reels.

  • Arma un video explicativo de 45 segundos con subtítulos animados a partir de este guion [pega el guion].

  • Convierte estos tres datos de la encuesta a clientes en un video corto con gráficos animados.

#Qué hace

hyperframes es la skill de entrada de HyperFrames, el marco de HeyGen que convierte HTML en video. Cada proyecto tiene un index.html raíz que declara las variables del video y monta una escena por cada archivo de compositions/, con los medios y las fuentes en assets/. El video nuevo sale por defecto en 1920 por 1080 a 30 cuadros por segundo, y la skill pide no agregar voz en off, subtítulos ni otro formato si nadie los pidió, salvo música y sonido en películas de más de 15 segundos que cuentan algo.

El núcleo es un contrato de 12 reglas que el comando npx hyperframes check --json controla y explica cómo arreglar: cada elemento con tiempo lleva class="clip", data-start y data-duration; cada escena registra una sola línea de tiempo pausada; nada de Date.now(), azar sin semilla, pedidos a internet ni temporizadores, porque todo lo visible tiene que depender del tiempo del video; las fuentes no incluidas se cargan desde assets/fonts/; los filtros sobre el material filmado pasan por media-treatment y no por CSS. Suma reglas para que el proyecto se pueda seguir editando: un id corto por escena, textos y colores de marca como variables, y el contenido dentro del 90 % del cuadro (los textos clave, dentro del 80 %).

El ciclo de trabajo es concreto. Para una película de más de 15 segundos propone primero tres historias distintas, solo en palabras, y escribe un SPEC.md donde cada adjetivo pasa a ser un número. Después de editar corre check, toma capturas con snapshot --review y deja un REVIEW.md. Solo renderiza cuando apruebas la vista previa. Trae guías por tipo de video (lanzamiento, explicativo, video musical, datos, vertical, 3D, personaje) y deriva a sus skills hermanas: hyperframes-cli, hyperframes-registry (unos 450 bloques listos, como transiciones, títulos y gráficos) y media-use (música, voz, avatar e imágenes).

#Cómo aplicarla a tu negocio

Caso de ejemplo: una pyme que vende un software de turnos para clínicas lanza una función nueva y necesita un video de 30 segundos para la web y un corte vertical para reels.

  1. Parte de lo real. Le pides: «Haz un video de lanzamiento de 30 segundos de la nueva agenda online, usando nuestra web». La skill captura el sitio con npx hyperframes capture <url> y usa tus pantallas, tu logo y tus colores en lugar de dibujarlos.
  2. Elige la historia. Te propone tres guiones distintos en texto (por ejemplo: el problema de las llamadas perdidas, un día de una recepcionista, el antes y después de la agenda). Eliges uno y la skill fija en el SPEC.md duraciones, movimientos y cortes.
  3. Revisa antes del render. Abres la vista previa, corriges una frase o un color, y recién ahí pides el MP4.
  4. Saca el vertical. Con la guía de clips verticales, reacomoda las escenas a 9:16 para reels y TikTok, con subtítulos animados si los pides.

Como todo queda en archivos, el mes siguiente cambias el texto y la captura de la función nueva y repites el render sin rehacer la animación.

#Cómo instalarla

HeyGen recomienda instalar el conjunto básico de skills, porque esta llama a sus hermanas (hyperframes-cli, hyperframes-registry, media-use).

  • Con una línea, para cualquier agente: npx skills add heygen-com/hyperframes --skill hyperframes -g instala solo esta; npx hyperframes skills update instala el conjunto básico completo, en la versión que coincide con el CLI.
  • Claude Code, como plugin: claude plugin marketplace add heygen-com/hyperframes y después claude plugin install hyperframes@hyperframes. Se llama con /hyperframes:hyperframes.
  • A mano: copia la carpeta skills/hyperframes/ completa (con references/, scripts/ y assets/) a ~/.claude/skills/ para Claude Code, o a ~/.agents/skills/ para Cursor, Codex y Gemini CLI (las tres leen esa carpeta).
  • App de Claude: comprime la carpeta en un ZIP y súbela desde Customize, Skills, «Upload a skill», sabiendo que el render necesita herramientas que el entorno de la app puede no tener.

Necesitas Node.js 22 o superior y FFmpeg. El SKILL.md completo (con la licencia y la fuente) está más abajo.

El SKILL.md completo

Skill original de HeyGen (Apache 2.0). Fuente: github.com/heygen-com/hyperframes/tree/main/skills/hyperframes.

El archivo original está en inglés y va tal cual, sin traducir. Lo de arriba de esta ficha es el resumen en español: qué hace, cómo aplicarla y cómo instalarla.

Leer el archivo (86 líneas)

Running the CLI: every npx hyperframes <command> here, the beacon included, means node <SKILLS>/hyperframes/scripts/plugin-cli.mjs <command> run in the project, where <SKILLS> holds this skill's folder (rules).

Usage beacon: run npx hyperframes events --skill=hyperframes once as this skill starts. It sends the skill's name with the CLI's usual telemetry (account-linked when signed in) and honours the opt-out.

HyperFrames

HyperFrames renders HTML to video by seeking a paused timeline frame by frame, so every render is identical; the person edits the same files in the Studio preview. Use it unless another framework or a plain browser recording is asked for. Run the CLI as npx hyperframes <command> (/hyperframes-cli); at the start of a new video, npx hyperframes usage --json reports the person's remaining plan usage (say "unknown" when it is).

index.html is the root composition: it declares the video's variables and mounts one scene per sub-composition from compositions/. Media and fonts live in assets/. A new project: npx hyperframes init <dir> --skip-skills --skill=hyperframes. A website as the subject: npx hyperframes capture <url> --json.

Build what was asked: no voiceover or captions and no new aspect ratio unless the request or the source calls for them; a film over ~15 s that tells something gets music and sound on its visible actions unless the person asked for silence (references/sound.md). A new video is 1920×1080 at 30 fps. A question, a hold or a loose idea gets an answer or a proposal, not an edit. Before hand-writing a named effect, transition, caption style, chart or card, search the registry (/hyperframes-registry). Any media (a presenter or avatar, footage, voice, music): /media-use, which says when HeyGen is needed and what is free. Show real things from their source (the product's screens, the brand's logo, a real place's imagery); draw only what has none.

Contract

npx hyperframes check --json reports most violations with a Fix: line.

  1. Every timed element has class="clip", data-start and data-duration in seconds, and shows over [start, start + duration): land an end state just before its end. data-track-index is only a Studio lane (clips that never overlap share one: references/format.md); paint order is z-index.
  2. The video's length is the root's data-duration: overrun is cut, an early end holds the last frame.
  3. Each composition builds one paused timeline and registers it after its tweens: window.__timelines["<data-composition-id>"] = tl.
  4. No Date.now(), unseeded Math.random(), network fetches, timers or repeat: -1. Everything visible is a function of timeline time.
  5. Never tween display, visibility or autoAlpha on a .clip; animate a child.
  6. No CSS transform on an element GSAP transforms; animate transforms, not left/top/width/height.
  7. Never play(), pause() or seek media. A <video> keeps its sound with data-has-audio="true", else it is muted; music, voice, replacement sound and J/L cuts are a separate <audio>.
  8. Every <audio> has an id or it is left out of the mix. No crossorigin on media. Time the <video>, never its wrapper.
  9. Only the bundled fonts (references/format.md) resolve by name; any other family needs @font-face to a file in assets/fonts/.
  10. Asset paths: assets/... (no leading /), in sub-compositions too.
  11. Grades, grain, blur and vignettes on footage go through npx hyperframes media-treatment (/media-use), never CSS filters.
  12. A scene file puts its styles, markup and scripts inside <template> and styles its root as #root. Intended overflow or overlap: data-layout-allow-overflow, data-layout-allow-overlap.

Keep it editable

Studio edits the tags you write.

  • One scene per sub-composition, short readable ids; the root holds only scene hosts, media and audio.
  • Timed elements are written in the HTML, not made by script, each with a stable id.
  • Copy, brand colours and swappable media are variables (data-var-text, data-var-src, var(--id)).
  • One kind per track: data-track-kind="graphics" on scene and overlay hosts, captions on the one host that holds every caption group.
  • Content stays inside the 90% action-safe box, captions and key text inside 80% title-safe; Studio draws the guides, never add them.

Work loop

  • An existing project: npx hyperframes timeline --json shows what plays when; before any change, npx hyperframes history --since mine --who <you> shows what the person changed since your last turn: build on it (/hyperframes-cli).
  • A new film over ~15 s that tells something: its type's file (below), then references/story.md: three storylines that are three different films, in words only; the person picks unless they asked for the file or no stops. Then SPEC.md (references/spec.md): every adjective as a number, the defaults banned by name, the shots by frame.
  • After your last edit: npx hyperframes check --json; fix every error, weigh every warning. A film: npx hyperframes snapshot --review and references/review.md, ending in REVIEW.md. A small edit: snapshot what you changed (--at, --range, --composition) and look before you reply.
  • Render when the person approves the Studio preview, asks for the file or asks for no stops, after your check and review pass; otherwise stop at the preview and wait.

The HyperFrames desktop app

Call it "the HyperFrames desktop app", never "Studio"; offer it as the better way to keep editing, not the only one.

  • Before the delivery render of a new video, open the Studio preview, even in no-questions mode, and say they can edit there. If preview printed "Keep editing by chatting with Framey", add that the Edit with Framey button opens the app.
  • After the delivery render, pass on its Framey line in your own words: offer to run the hyperframes open it names or give its download link; else say nothing (autonomous: the delivery note).
  • When any command says to run catch-up, run npx hyperframes catch-up [dir] before your next change; what it lists is their work, not a new request.

What to read

The requestRead
A video to copy or remake/reverse-any-video
An episode in a proven style/style-sampling
Another film's style on a new topicreferences/look.md
Launch, promo, featurereferences/types/launch.md
Explainer, tutorialreferences/types/explainer.md
Music videoreferences/types/music.md
Chart, numbersreferences/types/data.md
Vertical or social clipreferences/types/vertical.md
3D, simulation, generativereferences/types/3d.md
Character, mascot, story shortreferences/types/character.md
A real place or map; a website, PR or script; kinetic typereferences/film.md
Captions; cards over a talking headreferences/captions.md; references/overlays.md
Motion: eases, titles, cuts, punch-ins, speed, holdsreferences/motion.md
Scenes, attributes, media, lint; GSAP, Lottie, 3D, canvasreferences/format.md; references/runtimes.md

Decks: /slideshow. Remotion source: /remotion-to-hyperframes. Figma: /figma. Skills and the pinned CLI: references/skill-lifecycle.md.

Copiar el archivo tal cual
---
name: hyperframes
description: Read first to make, edit, animate or render any video with HyperFrames (HTML rendered to video), including launch films, explainers, music videos, charts, maps, vertical clips, title cards, captions and overlays. Holds the render contract, the work loop and what each kind of video needs.
---

<!--
Skill original: hyperframes, de HeyGen (https://github.com/heygen-com/hyperframes/tree/main/skills/hyperframes)
Licencia: Apache 2.0. Reproducida con atribución según los términos de la licencia.
Copia publicada en claura-ai.com/skills.
-->

**Running the CLI:** every `npx hyperframes <command>` here, the beacon included, means `node <SKILLS>/hyperframes/scripts/plugin-cli.mjs <command>` run in the project, where `<SKILLS>` holds this skill's folder ([rules](references/plugin-installation.md)).

**Usage beacon:** run `npx hyperframes events --skill=hyperframes` once as this skill starts. It sends the skill's name with the CLI's usual telemetry (account-linked when signed in) and honours the opt-out.

# HyperFrames

HyperFrames renders HTML to video by seeking a paused timeline frame by frame, so every render is identical; the person edits the same files in the Studio preview. Use it unless another framework or a plain browser recording is asked for. Run the CLI as `npx hyperframes <command>` (`/hyperframes-cli`); at the start of a new video, `npx hyperframes usage --json` reports the person's remaining plan usage (say "unknown" when it is).

`index.html` is the root composition: it declares the video's variables and mounts one scene per sub-composition from `compositions/`. Media and fonts live in `assets/`. A new project: `npx hyperframes init <dir> --skip-skills --skill=hyperframes`. A website as the subject: `npx hyperframes capture <url> --json`.

Build what was asked: no voiceover or captions and no new aspect ratio unless the request or the source calls for them; a film over ~15 s that tells something gets music and sound on its visible actions unless the person asked for silence (`references/sound.md`). A new video is 1920×1080 at 30 fps. A question, a hold or a loose idea gets an answer or a proposal, not an edit. Before hand-writing a named effect, transition, caption style, chart or card, search the registry (`/hyperframes-registry`). Any media (a presenter or avatar, footage, voice, music): `/media-use`, which says when HeyGen is needed and what is free. Show real things from their source (the product's screens, the brand's logo, a real place's imagery); draw only what has none.

## Contract

`npx hyperframes check --json` reports most violations with a `Fix:` line.

1. Every timed element has `class="clip"`, `data-start` and `data-duration` in seconds, and shows over `[start, start + duration)`: land an end state just before its end. `data-track-index` is only a Studio lane (clips that never overlap share one: `references/format.md`); paint order is `z-index`.
2. The video's length is the root's `data-duration`: overrun is cut, an early end holds the last frame.
3. Each composition builds one paused timeline and registers it after its tweens: `window.__timelines["<data-composition-id>"] = tl`.
4. No `Date.now()`, unseeded `Math.random()`, network fetches, timers or `repeat: -1`. Everything visible is a function of timeline time.
5. Never tween `display`, `visibility` or `autoAlpha` on a `.clip`; animate a child.
6. No CSS `transform` on an element GSAP transforms; animate transforms, not `left`/`top`/`width`/`height`.
7. Never `play()`, `pause()` or seek media. A `<video>` keeps its sound with `data-has-audio="true"`, else it is `muted`; music, voice, replacement sound and J/L cuts are a separate `<audio>`.
8. Every `<audio>` has an `id` or it is left out of the mix. No `crossorigin` on media. Time the `<video>`, never its wrapper.
9. Only the bundled fonts (`references/format.md`) resolve by name; any other family needs `@font-face` to a file in `assets/fonts/`.
10. Asset paths: `assets/...` (no leading `/`), in sub-compositions too.
11. Grades, grain, blur and vignettes on footage go through `npx hyperframes media-treatment` (`/media-use`), never CSS filters.
12. A scene file puts its styles, markup and scripts inside `<template>` and styles its root as `#root`. Intended overflow or overlap: `data-layout-allow-overflow`, `data-layout-allow-overlap`.

## Keep it editable

Studio edits the tags you write.

- One scene per sub-composition, short readable ids; the root holds only scene hosts, media and audio.
- Timed elements are written in the HTML, not made by script, each with a stable `id`.
- Copy, brand colours and swappable media are variables (`data-var-text`, `data-var-src`, `var(--id)`).
- One kind per track: `data-track-kind="graphics"` on scene and overlay hosts, `captions` on the one host that holds every caption group.
- Content stays inside the 90% action-safe box, captions and key text inside 80% title-safe; Studio draws the guides, never add them.

## Work loop

- An existing project: `npx hyperframes timeline --json` shows what plays when; before any change, `npx hyperframes history --since mine --who <you>` shows what the person changed since your last turn: build on it (`/hyperframes-cli`).
- A new film over ~15 s that tells something: its type's file (below), then `references/story.md`: three storylines that are three different films, in words only; the person picks unless they asked for the file or no stops. Then `SPEC.md` (`references/spec.md`): every adjective as a number, the defaults banned by name, the shots by frame.
- After your last edit: `npx hyperframes check --json`; fix every error, weigh every warning. A film: `npx hyperframes snapshot --review` and `references/review.md`, ending in `REVIEW.md`. A small edit: snapshot what you changed (`--at`, `--range`, `--composition`) and look before you reply.
- Render when the person approves the Studio preview, asks for the file or asks for no stops, after your check and review pass; otherwise stop at the preview and wait.

## The HyperFrames desktop app

Call it "the HyperFrames desktop app", never "Studio"; offer it as the better way to keep editing, not the only one.

- Before the delivery render of a new video, open the Studio preview, even in no-questions mode, and say they can edit there. If `preview` printed "Keep editing by chatting with Framey", add that the **Edit with Framey** button opens the app.
- After the delivery render, pass on its Framey line in your own words: offer to run the `hyperframes open` it names or give its download link; else say nothing (autonomous: the delivery note).
- When any command says to run `catch-up`, run `npx hyperframes catch-up [dir]` before your next change; what it lists is their work, not a new request.

## What to read

| The request | Read |
| --- | --- |
| A video to copy or remake | `/reverse-any-video` |
| An episode in a proven style | `/style-sampling` |
| Another film's style on a new topic | `references/look.md` |
| Launch, promo, feature | `references/types/launch.md` |
| Explainer, tutorial | `references/types/explainer.md` |
| Music video | `references/types/music.md` |
| Chart, numbers | `references/types/data.md` |
| Vertical or social clip | `references/types/vertical.md` |
| 3D, simulation, generative | `references/types/3d.md` |
| Character, mascot, story short | `references/types/character.md` |
| A real place or map; a website, PR or script; kinetic type | `references/film.md` |
| Captions; cards over a talking head | `references/captions.md`; `references/overlays.md` |
| Motion: eases, titles, cuts, punch-ins, speed, holds | `references/motion.md` |
| Scenes, attributes, media, lint; GSAP, Lottie, 3D, canvas | `references/format.md`; `references/runtimes.md` |

Decks: `/slideshow`. Remotion source: `/remotion-to-hyperframes`. Figma: `/figma`. Skills and the pinned CLI: `references/skill-lifecycle.md`.

Preguntas frecuentes sobre la skill hyperframes

¿Qué es la skill hyperframes y para qué sirve?

hyperframes es la skill principal de HyperFrames, el marco de video de código abierto de HeyGen (github.com/heygen-com/hyperframes, licencia Apache 2.0). Le enseña a Claude a armar un video como si fuera una página web: escenas en HTML, animaciones con una línea de tiempo pausada, y un render que recorre esa línea cuadro por cuadro para sacar un MP4. Como no hay nada al azar, el mismo proyecto da siempre el mismo video. Sirve para videos de lanzamiento, explicativos, gráficos con números, mapas, subtítulos animados, placas sobre un video de alguien hablando y clips verticales para redes.

¿Es gratis? ¿Qué parte se paga?

El marco, el CLI y la skill son gratis y de código abierto, y el render local a MP4 no cuesta nada. La aplicación de escritorio (HyperFrames Studio) también es gratis. Lo pago es el plan Pro de HyperFrames, que suma los modelos de video, imagen, voz y música de HeyGen, y dos skills del mismo paquete que lo requieren: la que reconstruye un video existente y la que produce episodios en cinco estilos ya probados. Ten en cuenta además que, al arrancar, la skill corre un comando que manda a HyperFrames el nombre de la skill junto con la telemetría habitual del CLI; respeta la opción de no enviar datos si la desactivas.

¿En qué se diferencia de un generador de video con IA o de editar en CapCut?

Un generador de video con IA inventa la imagen a partir de una descripción: es rápido, pero los textos, el logo y las cifras pueden salir deformados y cada intento es distinto. hyperframes escribe el video: los textos son exactos, el logo es tu archivo, las capturas de tu producto son reales y cualquier cambio se hace corrigiendo una línea. Frente a CapCut, la diferencia es que no arrastras nada en una línea de tiempo: le pides el cambio a Claude o lo tocas en la vista previa. Para tomas filmadas o personajes realistas sigue haciendo falta otra herramienta, y la skill deriva ese material a su compañera media-use. Más opciones en skills para video y reels.

¿Qué necesito instalado? ¿Funciona en la app de Claude?

Necesitas Node.js 22 o superior y FFmpeg, más un agente que corra comandos en tu computadora (Claude Code, Codex, Cursor o Gemini CLI). La skill descarga el CLI de HyperFrames por npm la primera vez. En la app de Claude puedes subirla como ZIP, pero el render depende de esas herramientas y de descargar el CLI, que el entorno de la app puede no tener, así que lo seguro es un agente de terminal. Las instrucciones están en inglés; pide los textos del video en español desde tu primer mensaje. Si prefieres trabajar con Remotion, mira la guía para editar reels con Claude Code y Remotion.

Más skills de Video

Para seguir con este tema

Elegí qué cookies usamos