no-slop-motion skill

A skill that walks an agent through ten gates, from customer pain to final render, so a launch film looks directed.

  • 92 ★ on the repository
  • Last commit
  • MIT
Skill

Extended access

Why this level

  • Scripts run locally and voice text goes to the voiceover service under your key.
All reasons and checks

ferndesk/no-slop-motion

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/no-slop-motion

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Scripts run locally and voice text goes to the voiceover service under your key.

This is third-party code. Review the repository files before installing.

What it does

no-slop-motion is a skill for launch and brand films. The agent works through ten gates: inputs, customer pain, script, voice, world and style frames, hero, animatic, build, sound and final QA. The next gate opens only after a person signs off the previous one. The repository holds a guide for each gate, templates for the brief, script and shot list, a HyperFrames starter project and scripts. The scripts record Cartesia voiceover takes, score them, split the voice in silences, join music on downbeats, lint banned patterns, detect pops and flashes, build contact sheets and export the film for social platforms.

Who it is for. Product teams and marketers who want an agent to build a product announcement with voice, music and sign-off at each step.

Good fit when

  • A launch film needs voiceover and music.
  • An AI-made video looks like a slideshow and keeps getting rejected.
  • A brand has a mascot or signature object to build the story around.
  • A video for LinkedIn or X is built from real customer words in call transcripts.

Not a fit when

  • You want a quick cut of clips with no script or sign-offs.
  • You have no voiceover service key and no music tool.
  • The video is made in an editor without HTML, so HyperFrames does not fit.

Example request

Make a launch video for our product for X. The site is acme.com, call transcripts are in ./calls, and I will name a reference separately.

Limitations

Needs Node 22 or newer, ffmpeg, Google Chrome or the HyperFrames headless shell, Python 3.10 or newer, a `CARTESIA_API_KEY` for voice and a music tool such as Suno or a licensed track. The skill builds on HyperFrames, whose skills install separately. The method needs several sign-off rounds, and story and style take the most time. Output quality depends on the inputs: without call transcripts and brand assets the customer pain and brand world gates come out generic.

How to disable. Delete the skill folder from the agent's skills directory. Film projects stay in their own folders and are untouched.

FAQ

How do I install no-slop-motion in Claude Code?

Run `npx skills add ferndesk/no-slop-motion`. By hand, copy `skills/no-slop-motion` into `~/.claude/skills/`. The HyperFrames skills come from `npx hyperframes skills`.

How does the skill differ from HyperFrames?

HyperFrames renders HTML to video. no-slop-motion adds the process: customer research, script, voice, style and checks, and leaves rendering to HyperFrames.

What does the agent show a person before the full build?

A one-page brief, a script, a voice, five style frames and an animatic. These sign-offs save rework at later gates.

Which tools are needed for voice and music?

Voice is recorded through Cartesia with the key in `CARTESIA_API_KEY`; ElevenLabs also works. Music comes from Suno or a licensed track.

Does the skill check the finished film?

Yes. Scripts look for banned patterns, pops and flashes and build contact sheets, and an independent reviewer agent watches every frame with sound on and muted.

Discussion

No questions or reviews yet.

Skills

A skill that makes your agent film a short product video: concept, edit, sound and render on Remotion or HyperFrames

SkillExtended access1.2KRepository stars

A skill that turns an article, README or website into a vertical HyperFrames promo video: brief and approval first, then assets, build and review

SkillExtended access483Repository stars

A kit for software product promo videos on HyperFrames: 101 themes, a transition atlas and a repetition check for storyboards

SkillExtended access255Repository stars

A Claude Code skill that builds a Remotion film of your product from its own components, tokens and logo, cut to music

SkillExtended access425Repository stars

A skill for promo films and kinetic type: idea and storyboard before the build, MP4 rendered through Chrome and ffmpeg, checked by six gates

SkillExtended access1.6KRepository stars

README in short

Agents left to defaults make the same film: a montage on a loud track, card grids and glowing gradients. The skill replaces those habits with a process where every decision rests on something real: customer pain, brand assets, the narrator breathing.

README badge

Original SKILL.md

---
name: no-slop-motion
description: Make a launch film, brand film, product announcement or promo video for a company that looks directed, not AI-generated. Use when someone asks for a launch video, brand video, sizzle, promo, product film or "a video for our brand/X/LinkedIn launch", or when an AI-made video "feels vibe coded", "looks like a slideshow", or keeps getting rejected. Covers the whole path: customer pain research, script, voiceover, world and style frames, optional mascot or hero object, animatic, HTML motion build (HyperFrames + GSAP), music, sound, QA and platform export.
---

# No-slop motion

A process for making a brand or launch film with an agent that people actually want to watch.

**The core idea: the agent's defaults are the slop.** When an agent picks something on its own (a track, a transition, a shadow, a font, a place to cut the voice), the result looks like every other AI video. What makes a film feel directed is that every choice comes from something real: the customer's own pain, the brand's own assets, the structure of the music, the narrator's breathing, or a named reference. So this skill is mostly about *where each decision comes from*, and the tooling is there to enforce it.

## What makes a film work

1. **Knowing the customer.** Who buys (the ICP), what hurts them in their own words, and the one thing only this product does (the USP). The film sells one before and after, not a feature list.
2. **One story thread.** Effect, then cause, then fix. For an invoicing tool: "why is the bank balance empty again?" → rewind → one invoice sat unsent for three weeks → the product arrives → it sends that invoice the moment work is done → what else you get → the balance fills up. A small mystery pulls people through a film. A feature tour doesn't.
3. **A world with its own rules.** A visual concept that comes from the story's feeling (for example a flickering 1970s office of paper forms and rubber stamps for the problem half, then clean brand colour when the product arrives). Every texture, colour and transition then has a reason.
4. **Brand motifs used freely.** Three to five things only this brand owns (a mascot, the logo mark, a signature button, a success animation, a sound). No mascot is fine: see [references/05-hero.md](references/05-hero.md).
5. **The voice sets the clock.** The narration is recorded first, in one continuous take per act. Picture is timed to the voice. Music comes last and is edited so it still sounds like one performance.
6. **Gates, not versions.** Each layer is locked before the next one is built, and each lock uses whatever the decision maker (the person who approves the film) can judge fastest: a page of text, five still frames, a rough animatic. Building story, style, voice and motion all at once is the slowest way to get there.

## Before you start

Check the tools in the first five minutes, not on day two:

| Need | Default | Notes |
| --- | --- | --- |
| Motion and render | [HyperFrames](https://hyperframes.heygen.com) (HTML + GSAP to MP4) | `npx hyperframes skills` installs its agent skills. Read `/hyperframes-core` before writing composition HTML. |
| Voice | [Cartesia](https://cartesia.ai) Sonic (`scripts/audio/tts-cartesia.ts`) | ElevenLabs works too. The key goes in an env var. Record a scratch voice in the first pass. |
| Music | [Suno](https://suno.com) or a licensed track | Once a track is liked, ask for continuations of it, not a fresh batch. |
| Transcription | faster-whisper or `npx hyperframes transcribe` | Word timestamps feed captions and voice splitting. |
| Audio and video tools | ffmpeg, Python 3 with numpy, librosa, soundfile | `pip install -r scripts/audio/requirements.txt` |
| Real UI | Access to the live product, marketing site and brand files | UI gets rebuilt as HTML from the product's real CSS and copy. |

Start the project from [starter/](starter/): a HyperFrames project with a film engine (one timeline, a timing table, actors, cameras, captions, sound cues). Copy it, then read `starter/README.md`.

## The gates

Work through them in order. A gate opens only when the decision maker has signed off the lock before it. Details and checklists are in the reference for each gate.

| Gate | What you make | Lock | Reference |
| --- | --- | --- | --- |
| 0. Inputs | Input pack: brand truth from production, product surfaces, claims list, platform and length, 2 to 4 references with what to take from each | Pack exists, `LOVES-HATES.md` started | [01-inputs.md](references/01-inputs.md) |
| 1. Pain and positioning | Pain bank from real customer words, one villain, 2 or 3 felt numbers, one-page brief | Brief approved | [02-story.md](references/02-story.md) |
| 2. Script | The film as spoken paragraphs per act | Words approved when read aloud | [02-story.md](references/02-story.md) |
| 3. Voice | Cast, record one take per act in 2 or 3 readings, score, split in silences | VO timeline. From here the voice sets the clock | [03-voice.md](references/03-voice.md) |

Report a problem