← Plugin catalog
Creativity

Sceneform AI Ads & Influencers

Sceneform v0.9.0

AI influencer & ad generator

Create a free account to save this plugin.

From the developer

Declared capabilities

  • Read
  • Write

Declared limitations

  • Unrelated request. Sceneform makes media and has nothing to add.
  • Text-only writing. The user asks for words, not media.
  • Sexual content, which Sceneform does not make.

Not specified in the available declarations: pricing & account requirements.

Onboarding example

Create an AI influencer for my skincare brand
Sources & dates

Developer declarations; not independently tested. Listing last checked Nov 17, 2026. Package observed Nov 16, 2026.

View package source →

Description

From the marketplace listing

Sceneform is an AI influencer and ad generator. Describe what you want in plain words and Sceneform starts making it. A card in the chat follows the job and shows the finished image, video, or audio, ready to download. What you can make: - AI influencers. Create a Persona from a description, or from photos you have permission to use, then cast it in images and UGC-style videos. - Ads. Static image ads and creator-style video ads built from your BrandKit's logo, colors, and product photos. A finished image ad can get new headlines, sizes, or languages. - Multi-slide carousels for social posts. - Images and short videos from a text prompt, or from an image you already have. - Voiceovers, music, sound effects, and lip sync. Build a BrandKit once from your website or a social profile, and every ad uses it. Schedule finished posts to TikTok, Instagram, YouTube, LinkedIn, or Facebook. Sceneform is for marketers, founders, and creators who make ads and social content. You need a Sceneform account. Videos take a few minutes to finish. Posting needs a social account connected in Sceneform.

Language: English · Automatically detected from descriptions.

Changes

No changes recorded yet. Follow this plugin to keep track of future changes.

First seen Nov 16, 2026. Last seen Nov 17, 2026.

Keywords

Files & skills

File archives

Plugin package43 files · 86.7 KBBrowse files →
Skill instructions
sceneform-ad-studio5.8 KB

View saved version →

---
name: sceneform-ad-studio
description: Create static ads, UGC and creator-style video ads with hooks, carousels, and new headlines, sizes, languages, or Reshoots of finished ads in Sceneform. Use when the user asks for an ad, ad creative, UGC or creator video for a brand or product, product promotion, campaign, A/B variants, or a swipeable post. NOT for a plain image or clip without an ad brief (sceneform-generate) or animated text and graphics (sceneform-motion).
license: MIT
metadata:
  author: sceneform
  version: "0.9.0"
---

# Ads and carousels

The ad tools write the detailed visual direction themselves. Your job is the
strategy they cannot invent: the brand facts, the angle, the audience, the
offer, and a concept sharp enough that the ad says one thing well. Read a
reference file by calling `load_skill` with this skill's name and the file's
`path`, or open it when this skill is installed locally.

## Steps

### 1. Ground the brand

Call `list_brand_kits`. Use a `ready` kit and its `product_ref` when the ad
sells one product; a product already carries its brand.

- No kit for the brand or product yet: load sceneform-brand and follow its
  build steps, then come back here.
- A kit still `building`, or a product whose `status` is `importing`, is
  not ready. With a card, say it is on its way and stop; otherwise check
  `get_brand_kit` on a later turn. A `failed` product can be imported again
  with `add_product`. An ad made from an unread brand or product comes out
  generic.
- No kit and no link: the concept carries every fact, and you offer to build
  a kit for the next ads.

Done when the ad has a ready kit or product, or the user chose to go without
one.

### 2. Choose format, count, and sizes

- Static image ad: `generate_static_ad`. A UGC or creator video, or any
  video ad: `generate_video_ad`, which picks the model and writes the
  script and shots itself. Swipeable post with a caption:
  `generate_carousel`.
- One ad unless the user asks for more. "A few", "some options", or "to
  test" means three to five concepts, each on a different angle.
- Static ad sizes: 4:5 for Instagram and Facebook feeds (the default), 9:16
  for Stories and Reels, 1:1 for LinkedIn and square placements, 16:9 for
  YouTube and the web. New sizes of a finished ad are variants, not new ads.
- Video ads run 4 to 30 seconds; 15 is the default and suits most feeds.
  They are 9:16 at 720p unless the user asks for another `aspect_ratio` or
  `resolution`. `ad_type` picks the style (UGC by default; unboxing, ASMR,
  TV spot, product showcase, tutorial, Japanese TV ad); only UGC, unboxing,
  ASMR, and tutorial take a hook or setting.
- Static ads and their variants render at 1K unless the user asks for 2K
  or 4K in `resolution`.

Done when you know each ad's tool, angle, and size.

### 3. Write the concept

Read [references/concepts.md](references/concepts.md) and write each
`concept` from its template: one angle, the audience, the message, the
proof, the offer, and the call to action, at most 1,200 characters. For a
video ad add a `setting_brief`: one line naming the place or situation.

Done when each concept passes the checklist in that file.

### 4. Pick the template, hook, or recipe

Read the matching catalog: `list_ad_templates` for static layouts,
`list_ad_hooks` for video openings, `list_carousel_recipes` for carousel
styles. Pick the one whose description fits the angle, and pass its id as
`template_id`, `hook_id`, or `recipe_id`. Leave it out to let the tool pick.

- A person in the ad: a ready `persona_ref` from `list_personas` in
  `reference_refs` (up to 3 images for static, 10 for video), chosen by the
  casting rules in the concepts file. The product stays in `product_ref`.
- An `element_ref` from `list_elements` or an image `asset_ref` also goes in
  `reference_refs`. A carousel takes a `persona_ref` as a recurring
  character. For a new reusable AI influencer, use sceneform-brand first.

### 5. Generate

`generate_static_ad

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 3 — file references are metadata, not package contents.

sceneform-brand6.58 KB

View saved version →

---
name: sceneform-brand
description: Read, build, and edit Sceneform BrandKits and their products, and create or edit Personas, reusable AI characters or influencers made from a description or photos. Use when the user mentions their brand, website, products, colors, voice, a saved character, or wants content in their style. NOT for one-off media (sceneform-generate), ads (sceneform-ad-studio), or posting (sceneform-publish).
license: MIT
metadata:
  author: sceneform
  version: "0.9.0"
---

# BrandKits and Personas

A BrandKit is a brand's saved identity: name, website, summary, audience,
voice, colors, logo, and products. A Persona is a saved, reusable character
with a portrait and an optional default voice. Both exist so every
generation reads as the same brand and the same person. Read a reference
file by calling `load_skill` with this skill's name and the file's `path`,
or open it when this skill is installed locally.

## Read

All reads are free.

- `list_brand_kits`: every kit with `brand_kit_ref`, `name`, `status`,
  `summary`, `audience_summary`, `voice`, `voice_exemplars`, `colors`, and
  `products` (each with a `product_ref`, `name`, and `status`: `importing`
  while its page is read, then `ready` or `failed`).
- `get_brand_kit`: one kit, with `status` `building`, `ready`, or `failed`,
  and once ready its `logo_url`. `get_persona` reads one Persona the same
  way. Reads never show a card. When the user wants to review a Persona
  and a BrandKit together, call `show_card` with just those two refs.
- `list_personas`: every Persona with `persona_ref`, `name`, `status`,
  `has_portrait`, and `has_voice`. Only a `ready` one goes into a
  generation.
- `list_elements`: saved Elements (reusable characters, locations, and
  props) with `element_ref`, `name`, `category`, `description`, and
  `image_count`. Pass `category` to list one kind.

## Use a brand in a generation

1. Read the kit before writing a prompt or concept for that brand. Turn its
   `summary`, `audience_summary`, `voice`, and `colors` into plain words in
   the prompt ("deep forest green and cream, calm and plain-spoken"), and
   quote a `voice_exemplars` line when writing copy.
2. Facts come from the kit or the user: claims, prices, label text, colors.
   A field the kit leaves empty stays out of the prompt.
3. A Persona, product, or Element goes into an image or video through
   `reference_refs` (`generate_image`, `generate_video`, `reshoot_video`,
   standalone ads). A Persona
   also works as `start_frame_ref` to animate its portrait, or as an
   avatar's `image_ref`.
4. Refs stay inside tool arguments. In prose, use the brand, product, and
   Persona names.

## Build and edit a BrandKit

These writes are free and save at once.

- `build_brand_kit_from_source` builds a kit from the brand's website or a
  social handle; prefer it whenever the user names one. A bare domain or
  "my brand" is a brand source. A product page link (`/products/...`) is a
  product: `add_product` into the brand's kit, and when there is no kit yet,
  build one from that site's domain first, then add the product to it. When
  the user gives only a brand name, ask for the link rather than guessing a
  domain.
- The build returns status `building`. With `card_offered: true` the host
  was given a card that follows it; otherwise, or when the user cannot see
  the card, check `get_brand_kit` on a later turn. Content made
  before the kit is `ready` ignores it. A `failed` kit can be built again
  from the same source.
- `create_brand_kit` makes a kit from an identity the user spells out:
  `name`, `website` (or null), `brand_summary`, and optionally the audience,
  voice, `voice_lines`, colors, and a logo `logo_ref`.
- `update_brand_kit` changes only the fields you pass; null clears a field.
  Logos, documents, and competitors stay on the BrandKits page.
- `add_product` imports a product from its page: `brand_kit_ref`, `url`,
  and `kind` `"app"` for software. It returns a `product_

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 1 — file references are metadata, not package contents.

sceneform-generate11 KB

View saved version →

---
name: sceneform-generate
description: Generate images, video, speech, music, sound effects, voice changes, lip sync, and talking avatars with Sceneform, and edit, animate, upscale, or reframe existing media. Use when the user wants any of these from a brief, a reference, or a file they already have, including a viral remix, a narrated storybook explainer, or a restyled clip. NOT for ads or carousels (sceneform-ad-studio), BrandKits or Personas (sceneform-brand), animated text and graphics (sceneform-motion), or posting (sceneform-publish).
license: MIT
metadata:
  author: sceneform
  version: "0.9.0"
---

# Generate with Sceneform

You are the art director and the prompt writer. Sceneform sends your prompt
to the model exactly as you write it, so the result is only as specific as
the prompt. The craft lives in reference files: read one by calling
`load_skill` with this skill's name and the file's `path`, or open it when
this skill is installed locally.

## Steps

### 1. Read the brief

Name the kind (image, video, speech, music, and so on), the deliverable,
where it will appear, and how many. Take every value the user gave. Fill the
rest with defaults: one output, the aspect ratio of the destination (the
craft references list them), and the style the subject and brand imply.

Ask only about a gap that blocks a useful result, and bundle up to three
short questions in one message:

- the subject, when nothing describes it;
- whose face, when a specific real person must appear and there is no photo
  or Persona;
- what to change, when the user rejects a result without saying why.

Model, lens, lighting, and prompt wording are your decisions. When the user
says "just make it", "go ahead", or similar, ask nothing: state the plan in
one sentence and generate.

When the work is for the user's brand or product ("our", "my shop", a
product name), read it first with `list_brand_kits` and use its facts.
Claims, prices, colors, and label text come only from the user or the kit.
A UGC or creator video, or any ad for a brand or product, belongs to
sceneform-ad-studio: `generate_video_ad` writes the script and shots from
your concept, which beats a hand-written video prompt.

Done when every value has a source (the user said it, or you chose the
default) and you can state the plan in one sentence.

### 2. Load the craft for the kind

| The request | Read |
| --- | --- |
| An image, an image edit, upscale, background removal, reframe | [references/image.md](references/image.md) |
| A video clip, an animated image, a Reshoot of a clip | [references/video.md](references/video.md) |
| Speech, music, sound effects, voice change, lip sync, avatar | [references/audio.md](references/audio.md) |
| A viral remix, a storybook explainer, a restyled clip | [references/short-videos.md](references/short-videos.md) |
| Working from the `sceneform` CLI | [references/cli.md](references/cli.md) |

Then pick the model by the job. A model the user names wins; otherwise start
here:

| The job | Model |
| --- | --- |
| An ad, a UGC or creator video, a product demo | Not `generate_video`: sceneform-ad-studio, `generate_video_ad` |
| Any other video: people talking, dialogue and sound, several shots, references | Seedance 2.5 (the default) |
| Trying a video idea cheaply | Seedance 2.5 at `resolution: "draft"`, then `finish_draft` |
| Fast physical action: sports, dance, stunts | Kling 4.0 |
| A clip that must follow reference clips closely, or 4K | Gemini Omni Flash 1.1 |
| Many references at once, or copying a reference clip's motion | H3 Max |
| An image with text, a design, an ad, a product photo, or an edit | GPT Image 2.6 Sunburst (the default) |
| A photoreal scene, an illustration or cartoon character, or many reference images combined | Nano Banana Pro |
| An edit of a real photo that must keep the person's face | Seedream 5.0 Pro |
| A candid phone photo of a person, with no product or text in frame | Soul 2.0 |

For a job the table does not cover, [re

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 27 — file references are metadata, not package contents.

sceneform-motion13 KB

View saved version →

---
name: sceneform-motion
description: Make and edit motion videos in Sceneform, animated text and graphics built with the user's own images, clips, or audio. Use when the user wants an announcement, product feature, stat, quote, captioned clip, or short explainer as an animated video, or wants a finished one changed. NOT for camera footage or people on screen (sceneform-generate) or a creator-style video ad (sceneform-ad-studio).
license: MIT
metadata:
  author: sceneform
  version: "0.9.0"
---

# Motion video

You are the motion designer. Your model plans the film, writes the editing
code, inspects rendered frames, and refines the result. Sceneform supplies
the motion engine, storage, and renders. Blank project creation is free;
engine time (1 credit a minute) and full-quality exports cost credits. Read
a reference file by calling `load_skill` with this skill's name and the
file's `path`, or open it when this skill is installed locally.

## Agree the film first

A film built before the brief is agreed is rebuilt after it. Your first
message back is the storyboard, not a film: a launch film built straight
from the request was rebuilt in a new style a round later, and a remake
whose voices and music the user never chose was rebuilt again. Spend the
first round on words and a few stills, which cost minutes of session time,
and only then on renders and exports.

1. Brief. Ask only for what the request leaves open: a six-second
   announcement from a BrandKit needs no questions. For a launch or brand
   film, or anything over 15 seconds, settle these before building:
   - where it will be posted and who sees it (a first post on X and a
     landing-page hero are different films);
   - the one sentence the viewer should leave with;
   - whether it covers the whole product or one tool;
   - three or four reference videos the user likes, and one they do not.
     These are the most useful input you can get: read their shot lengths
     and pacing, and cut to the same numbers. Without any, choose a look
     from [references/looks.md](references/looks.md) and name it;
   - which assets are in bounds (their site, product shots, BrandKit) and
     what is off limits (labels, mockups, footage that reads as stock);
   - for a film "like the last one", what carries over (the look, the
     narrator, the music family) and what is new, since the user means its
     quality, not its scenes.
2. Storyboard. Before the first render, show the user a plan and get their
   approval: the look in two sentences, a beat sheet with timings, the
   on-screen copy read as one sentence from the first beat to the last,
   the narrator and the music direction by name, and one preview frame of
   the first beat built from real assets (`preview_motion_project` with
   `times_ms` once the assets are placed). Fix the subject, the scope, the
   copy, the voices and the chrome here, where a change is a sentence.
3. Build to the approved storyboard. Iterate with preview sheets and
   `render_motion_draft`; export once, on sign-off. An export's price grows
   with the film's length (`get_motion_review` returns it as
   `export_credits`), and `export_motion_draft` is refused until the user
   could see the version as a draft. A film needs one export per
   approved milestone, not one per idea.
4. Keep a plan file (beat sheet, decisions, asset ids, timings, the values
   each change replaced) in the user's workspace while the job runs, so a
   compacted context, a handoff or a revert starts from it rather than
   from the chat.

## Make one

1. Settle the length and the aspect ratio. Lengths are 6, 10, 15, 30, or 60
   seconds; aspect ratios are 16:9, 9:16, or 1:1. When the user does not
   say, use 15 seconds and the shape of where it will be posted: 9:16 for
   reels and stories, 1:1 for feeds, 16:9 for websites and YouTube.
2. Gather inputs from `search_library`, `open_upload`, or `import_media`.
   Read the brand's BrandKit for its colors, typography, and voice.

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 3 — file references are metadata, not package contents.

sceneform-publish8.97 KB

View saved version →

---
name: sceneform-publish
description: Schedule, publish, and manage images, videos, and carousels on the user's connected TikTok, Instagram, YouTube, LinkedIn, Facebook, or X accounts, plus text-only X posts. Use when the user wants to post, schedule, or share content, see what is scheduled, where a post went live, or how posts did, or reschedule, edit, cancel, publish now, or retry a post. NOT for making the asset (sceneform-generate) or reading brand identity (sceneform-brand).
license: MIT
metadata:
  author: sceneform
  version: "0.9.0"
---

# Publish with Sceneform

`schedule_post` sends one finished asset or carousel to one connected social
account, either as soon as possible or at a set time. Posts go out through
the user's own account and appear on the Schedule page in the Sceneform
dashboard. The other tools here read the connected accounts and the
Schedule, and change posts that are already on it. For a text-only X post,
omit both media refs. X supports up to four images or one MP4 video (up to
512 MB in Sceneform), with up to 280 weighted characters; links count as 23
characters. The X account's audience settings apply. An X post can carry a
first comment: a reply from the same account posted right after it goes
live, the usual place for a link, since X shows posts without one to more
people. X analytics and threads are not supported. Each X post costs
Sceneform credits, more when its text has a link, and a first comment is
priced as a second post; the approval shows the exact cost. A post that
fails before X receives it is refunded, and so is a first comment X
refuses.

## Workflow

1. Get the asset. Use an `asset_ref` from a finished operation
   (`wait_operations` or `get_operation`, `status: "succeeded"`) or from
   `search_library`. The asset must be complete; a pending output cannot be
   scheduled. To post a whole carousel, pass its `carousel_ref` instead of
   `asset_ref` once `get_carousel` reports it `ready`: every slide goes out
   in order as one multi-image post. YouTube and Facebook take videos only,
   and each other platform caps the slide count; `schedule_post` names the
   limit when a carousel is too long. Skip this step for a text-only X post.
2. Read the account. `list_social_accounts` gives, for each connected
   account, `privacy_options` (who a post can be shown to: `public`,
   `followers`, `friends`, `private`), the TikTok `interactions` it can turn
   on (`comments`, `duet`, `stitch`), and for Facebook the `pages` it can
   post to.
3. Ask the user what only they decide. For TikTok, ask who can see the post
   (one of its `privacy_options`) and whether to allow comments, and duet and
   stitch for a video. TikTok requires the user's own choice: never pick a
   privacy for them, and leave every interaction off unless they turn it on.
   Ask once per conversation: when the user has already said, for example
   "public, comments on", reuse that for their later TikTok posts and say so
   with each quote, unless they change it. `private` makes a test post only
   the user sees, on TikTok and YouTube.
4. Write the caption in full, in the brand's voice (read the brand kit first
   with sceneform-brand when there is one). Add a short `title` only for
   platforms that show one, such as YouTube.
   - The first line carries the hook: feeds cut the caption after a line or
     two, so the reason to stop scrolling comes first.
   - One idea, then one call to action ("Link in bio", "Save this for
     later").
   - Three to five specific hashtags at the end on TikTok and Instagram;
     none or one or two on LinkedIn and Facebook.
   - Facts, prices, and claims only from the user or the BrandKit.
   - A paid partnership or a sponsored post says so in the caption.
5. Quote. Call `schedule_post` with `asset_ref` or `carousel_ref`,
   `platform` (`tiktok`, `instagram`, `youtube`, `linkedin`, `facebook`, `x`),
   `caption`, `privacy` (required for TikTok; defaults to `public`
   elsewhere), `interaction

Preview of the first 4,000 characters. Read full instructions →

Release notes

First release. Create AI influencers (Personas) and cast them in UGC-style video ads, make static image ads and carousels from your BrandKit, generate images, video, voiceovers, music, and lip sync, and schedule finished posts to your connected social accounts. Progress and results show in a Sceneform card in the chat.

Declared in the saved package. Remote tools may change independently.

Package details

Publisher declarations from the archived package. These are separate from our research and the live service's terms.

Package license
MIT
Package author
Sceneform
Keywords
See keywords
Review scenarios
5 positive · 3 negativeScenario counts, not test results.

Declared capabilities

  • Read
  • Write

Package observed Nov 16, 2026.

Technical details
First seen
Nov 16, 2026 · 18:00 UTC
Last seen
Nov 17, 2026 · 00:00 UTC
Collection status
Collected

plugin_asdk_app_6af20aca7d008191b1c8506b8f187772

Download plugin data (JSON)

Plugin details: served from cache. Prepared . Refreshed when plugin data changes or after 6 hours. Your account and Follow status are live.