add_product
ChatGPTImport a product into a BrandKit from a public product page, app store listing, or app site. Free. It imports in the background; wait until list_brand_kits shows it before using it in an ad.
analyze_media
ChatGPTReport what an image, video, or audio asset contains; audio returns a transcript. compare_ref puts a second asset beside it.
build_brand_kit_from_source
ChatGPTBuild a BrandKit from a website, Instagram handle, or TikTok handle the user names. load_skill sceneform-brand first.
cancel_scheduled_post
ChatGPTUnschedule a waiting or held post, or dismiss a failed one or a TikTok draft. A live post is deleted on the platform, not here.
change_voice
ChatGPTRe-voice a finished recording (audio_ref) in another voice (voice_ref), keeping its timing and emotion. Charges credits. load_skill sceneform-generate first.
check_motion_project
ChatGPTList likely mistakes in the draft, such as cut-off text, loudness, or hard cuts, and the export time estimate. Costs session time.
confirm_actions
ChatGPTRun the quotes the user agreed to, by action_ref; several are one call. Confirming a ref again never charges or posts twice, so retry a lost confirmation with the same ref.
create_brand_kit
ChatGPTCreate a BrandKit from an identity the user spells out. Free.
create_folder
ChatGPTCreate a library folder and get its folder_ref. Check list_folders first so an existing one is reused.
create_motion_project
ChatGPTCreate a blank motion project that you build with the editing tools; Sceneform makes no video for it. Free. load_skill sceneform-motion first.
create_project
ChatGPTCreate a Project and get its project_ref. Check list_projects first so an existing one is reused.
create_storybook
ChatGPTMake a narrated, illustrated explainer video from a topic or brief. Charges credits. load_skill sceneform-generate first.
create_viral_remix
ChatGPTRemake a viral short video with the user's Persona, product, and message. Charges credits. load_skill sceneform-generate first.
cut_video
ChatGPTCut up to 60 seconds of a video into its own asset. Free.
duplicate_motion_project
ChatGPTCopy a motion video's current draft into a new project, for a second cut or an A/B. Free.
edit_motion_project
ChatGPTChange the draft with JavaScript that returns editing actions against the revision you read; a failed batch changes nothing. Costs session time.
export_motion_draft
ChatGPTRender a motion video at full quality, once the user has seen this draft version from render_motion_draft. Charges credits.
fetch
ChatGPTFetch one media item by asset_ref: a library item from search, a generation output, or an upload.
finish_draft
ChatGPTRender a 480p Seedance 2.5 draft again in 1080p, keeping the same take. A draft finishes once, within 14 days; after that this makes a new 1080p video instead. Charges credits.
generate_ad_variants
ChatGPTMake new headlines, sizes, or languages of a finished image ad. Charges credits. load_skill sceneform-ad-studio first.
generate_avatar
ChatGPTMake a talking avatar: the face in image_ref, an image or a ready persona_ref, speaks audio_ref. Charges credits. load_skill sceneform-generate first.
generate_carousel
ChatGPTMake a carousel from a concept. Charges up to the quoted credits and refunds the rest. load_skill sceneform-ad-studio first.
generate_image
ChatGPTMake images from a prompt. A saved Persona, product, or Element goes in reference_refs: its persona_ref from list_personas, product_ref from list_brand_kits, or element_ref from list_elements. Variations of one prompt are amount. Charges credits. Before writing the first prompt, load_skill sceneform-generate and read the option's prompt_guide from query_generation_options action get.
generate_music
ChatGPTMake a music track from a description; settings.duration sets the length. Charges credits. load_skill sceneform-generate first.
generate_persona
ChatGPTCreate a Persona, a reusable AI influencer, from a description or 1 to 5 photos; pass photos only once the user confirms they may use every face in them. Charges credits. load_skill sceneform-brand first.
generate_sound_effect
ChatGPTMake a sound effect from a description. Charges credits. load_skill sceneform-generate first.
generate_speech
ChatGPTTurn a script into a voiceover, in voice_ref from query_generation_options action voices. Charges credits. load_skill sceneform-generate first.
generate_static_ad
ChatGPTMake a static image ad from a template, BrandKit, and product. Charges credits. load_skill sceneform-ad-studio first.
generate_video
ChatGPTMake a video from a prompt, optionally from a start_frame_ref and end_frame_ref or with reference_refs (assets, Personas, products, Elements). Variations of one prompt are amount. Charges credits. Before writing the first prompt, load_skill sceneform-generate and read the option's prompt_guide from query_generation_options action get.
generate_video_ad
ChatGPTMake a UGC-style video ad of 4 to 30 seconds from a concept, BrandKit, and product. Charges credits. load_skill sceneform-ad-studio first.
get_action
ChatGPTRead a quote by action_ref without confirming it. The Sceneform card calls it.
get_asset_url
ChatGPTGet the download URL for an asset_ref.
get_brand_kit
ChatGPTRead a BrandKit by brand_kit_ref, including whether it is still building.
get_carousel
ChatGPTRead a carousel's status, caption, and slides by carousel_ref.
get_credit_balance
ChatGPTRead the account's available credits and plan; check it before proposing paid work.
get_motion_review
ChatGPTRead a motion video's editing state and session time left; call it before editing and whenever a call says the draft moved. Editing costs engine time, charged as the calls use it: 1 credit for each minute started, at most 25 credits for a session's 25 minutes. A video's first session opens with the first editing call; once its time is used up, open_motion_session asks the user to approve another.
get_motion_video
ChatGPTRead a motion video by motion_video_ref: its status, its video once rendered, and with motion_operation_ref an export's progress. timeout_seconds waits for a render or export to settle.
get_operation
ChatGPTRead a generation operation by operation_ref: its status, credits, and outputs with URLs.
get_persona
ChatGPTRead a Persona by persona_ref: its status, portrait, and voice.
get_short_video
ChatGPTRead a viral remix, storybook, or restyle by short_video_ref: its status and, once ready, its video.
import_media
ChatGPTImport an image, video, or audio file from a public https link the user pasted, and get its asset_ref.
import_motion_asset
ChatGPTAdd a Sceneform asset, a public file, or a Google Fonts family to the draft. Costs session time.
inspect_motion_project
ChatGPTRead the draft's project document, or one layer or scene, optionally at a film time. Costs session time.
lip_sync
ChatGPTSync the mouth in a video (video_ref) to an audio track (audio_ref). Charges credits. load_skill sceneform-generate first.
list_ad_hooks
ChatGPTList Ad Studio video hooks to pass as hook_id to generate_video_ad.
list_ad_templates
ChatGPTList Ad Studio static ad templates to pass as template_id to generate_static_ad.
list_brand_kits
ChatGPTList the user's BrandKits with their identity, voice, colors, and products, and their brand_kit_ref and product_ref values. Write brand copy in the kit's voice and exemplar lines.
list_carousel_recipes
ChatGPTList carousel recipes to pass as recipe_id to generate_carousel.
list_credit_transactions
ChatGPTList the newest credit charges, grants, and refunds with the balance after each. Use it when the user asks where credits went or whether a refund landed.
list_elements
ChatGPTList the user's saved Elements (characters, locations, and props) with their element_ref.
list_folders
ChatGPTList the user's library folders, including each Project's folder and Favorites, with their folder_ref.
list_operations
ChatGPTList what this session started recently, with each ref. After a timeout, find the started work here instead of starting it again.
list_personas
ChatGPTList the user's Personas, their AI influencers and recurring characters, with status and persona_ref.
list_projects
ChatGPTList the user's Projects with their project_ref, to put new work in the one the user names.
list_restyle_styles
ChatGPTList Restyle styles to pass as style_id to restyle_video.
list_scheduled_posts
ChatGPTList the posts on the user's Schedule, by default from 7 days back to 30 days ahead, with their status and, once posted, their link and metrics.
list_social_accounts
ChatGPTList the user's connected social accounts with their visibility options, allowed TikTok interactions, and available Facebook Pages.
load_skill
ChatGPTLoad the Sceneform skill for a task before starting it: sceneform-ad-studio, sceneform-brand, sceneform-generate, sceneform-motion, sceneform-publish. Without a name it lists when to use each; with a path it reads a file the skill links.
open_home
ChatGPTShow the Sceneform home in the Sceneform view: tools, use cases, and the newest generations.
open_library
ChatGPTShow the user's Sceneform library, newest first, in the Sceneform view.
open_motion_session
ChatGPTOpen another edit session once a motion video's session time is used up; the first one opens on its own. Charges credits.
open_upload
ChatGPTLet the user upload a file from their device, including one attached in chat that came without a download_url. In a host that shows no card, follow the result's without_card.
preview_motion_project
ChatGPTRender the draft's frames as contact sheets so you can see the film. Costs session time.
publish_scheduled_post_now
ChatGPTPublish a scheduled post now, or retry a failed one, when the user asks for that. Check it later with list_scheduled_posts.
query_generation_options
ChatGPTFind a model for a generation: recommend, search, list, or get an option (get returns its prompt_guide), or list voices.
read_motion_reference
ChatGPTRead the motion engine's reference, the project and action schemas, or the shader and sound libraries. The engine document is free; the rest cost session time.
reframe_image
ChatGPTExtend one image, passed in reference_refs, to the aspect ratio in settings.aspect_ratio. Charges credits. load_skill sceneform-generate first.
remove_background
ChatGPTRemove the background from one image, passed in reference_refs. Charges credits. load_skill sceneform-generate first.
render_motion_draft
ChatGPTRender the draft at draft quality and show it on the video's page, once a turn's edits are done. Costs session time.
reshoot_video
ChatGPTChange who appears, what they wear or hold, the place, or the light in a finished video, keeping its motion, timing, and speech. Charges credits. load_skill sceneform-generate first, or sceneform-ad-studio for the user's own video ad.
restyle_video
ChatGPTTurn a 3 to 60 second clip into a style from list_restyle_styles, keeping its audio. Charges credits. load_skill sceneform-generate first.
schedule_post
ChatGPTSchedule or publish an image, video, or carousel post, or a text-only X post, through the user's connected account once they approve its quote. load_skill sceneform-publish first.
search
ChatGPTSearch the user's Sceneform media library. Each result id is an asset_ref that fetch, the generation tools, and analyze_media accept.
search_library
ChatGPTSearch the user's media library by what the media shows, or list the newest without a query. Returns asset_refs.
show_card
ChatGPTShow the user one card with up to 12 results they want to review now, by ref. Put only what belongs together on it, such as a before and after or the variants of one request.
update_brand_kit
ChatGPTChange a BrandKit's written identity or colors; only the fields you pass change. Free.
update_persona
ChatGPTRename a Persona or set its default voice. Free.
update_scheduled_post
ChatGPTMove a scheduled post to a new time, or replace its caption or title, when the user asks for that change. Pass only what changes.
upload_asset
ChatGPTGet a one-time URL to PUT a local file's bytes to, with the returned headers, and the asset_ref it becomes. For hosts that can run commands or send requests.
upscale_image
ChatGPTUpscale one image, passed in reference_refs. Charges credits. load_skill sceneform-generate first.
view_video_frames
ChatGPTSee a video's frames as timestamped contact sheets, from an asset or a public video link, to read it shot by shot or find a cut.
wait_operations
ChatGPTWait up to 25 seconds for operations to finish and return their outputs with URLs; call again for any still running.