account_status
ChatGPTThe selected Voyager workspace, its plan and available credits. Read-only.
add_cast_reference
ChatGPTAdd an image you own to a Cast card as a reference.
cancel_job
ChatGPTAsk to stop a queued or running image, video or speech job. Best effort: work the provider already did can still be charged, and a finished job is unchanged.
create_upload
ChatGPTFor a local file on disk (CLI agents such as Claude Code or Codex): a 15-minute signed PUT URL for ONE image, video clip or audio file the user explicitly supplied. Run the returned curl command, then call import_reference with the upload_id to use it as a reference.
delete_asset
ChatGPTPermanently delete one of your imported references or generated results now.
delete_cast
ChatGPTDelete a Cast card and the reference images no other card uses. Jobs already started keep theirs.
estimate_generation
ChatGPTOnly when the user asks what it will cost, gives a budget, or the job is unusually large: estimate the model and settings without starting or holding anything. A plain request generates directly. Omit model for the operation's default. Estimate only the model you will run, never each candidate. Use the same model and parameters for generation. A hold estimate is not the final charge.
generate_3d
ChatGPTMake a textured 3D model from text or one owned image, or retexture a finished 3D job. Spends credits. Returns a job; results include GLB and other downloadable formats. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-3d') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_cast
ChatGPTGenerate reference images from a description and save them as a new Cast card when the job finishes. Uses image credits; poll the returned job with get_job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-image') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_image
ChatGPTOmit model for the default; the result names the model used. Generate images with the exact selected catalog model. Spends credits from the selected workspace. Optionally pass owned reference images (file_…) and saved Cast cards. Returns a job; poll get_job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-image') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_music
ChatGPTOmit model for the default; the result names the model used. Generate music with the exact selected catalog model and supported settings. Spends credits. Returns a job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-audio') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_segmentation
ChatGPTSelect objects in an owned image or video and return them as files: one black-and-white mask per object (white = object), cut-outs, a preview, per-object mask videos for a video, or the raw mask data (JSON). Select by prompt (text, SAM 3 and SAM 3.1), points or boxes in the source's pixels. mode picks the result: segment-image (masks, cut-outs; the default for an image), segment-image-data, segment-video (the rendered video), segment-video-data (mask videos per object; the default for a video), segment-everything (SAM 2, every object, no selection), image-embedding. Mask file_ ids feed an edit's mask or a video edit's mask video. Spends credits. Returns a job; poll get_job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-image') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_sound_effect
ChatGPTOmit model for the default; the result names the model used. Generate sound effects with the exact selected catalog model and supported settings. Spends credits. Returns a job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-audio') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_speech
ChatGPTSpeak the exact text. Omit model and voice_id for the default (Eleven v4, voice Sarah); otherwise a voice from list_voices that is compatible with the model (a premade voice, or your own existing voice clone where your plan includes it). Spends credits. Returns a job; poll get_job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-audio') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
generate_video
ChatGPTOmit model for the default; the result names the model used. Generate video with the exact selected catalog model and its supported settings. Spends credits. Pass owned file_ ids under the catalog's reference fields in parameters, or use Cast. Returns a job; poll get_job. Before generation, load_skill('voyager-cloud') and load_skill('voyager-cloud-video') unless already loaded. Waits up to 20 seconds. If still active, keep calling get_job(wait_seconds=25), or poll every 5–10 seconds, until terminal; do not end your turn while running. Show the result. Stop on cancelled/failed or unknown without poll_after_seconds; report uncertainty, never resubmit.
get_asset
ChatGPTA reference or result by id, with fresh 10-minute view and download links.
get_brand_kit
ChatGPTA brand kit's colors, typefaces and logos (read-only). Use them in prompts, and pass a logo's file_… id as a reference image when it should appear.
get_cast
ChatGPTRead one Cast card by id or slug, with fresh links to its reference images.
get_job
ChatGPTWait up to 25 seconds for a job, returning early when terminal, with fresh result links and charge. Call again until succeeded/failed; do not end your turn while running. Poll about every 5–10 seconds without bounded wait. Show terminal results. Stop on cancelled or unknown with no poll_after_seconds; report uncertainty and use Refresh later, never resubmit.
import_reference
ChatGPTImport ONE image, video clip or audio file the user explicitly supplied (an attached file, an https URL, or the upload_id of a local file uploaded with create_upload) as a private reference for generation. Kept 7 days unless saved into a Cast card.
list_brand_kits
ChatGPTBrand kits in the selected workspace (read-only).
list_cast
ChatGPTList this workspace's Cast: saved characters, locations and props with their reference images.
list_jobs
ChatGPTYour recent Voyager jobs in this workspace, newest first.
list_models
ChatGPTWithout query or model: Voyager's short recommended list per operation, one marked default and what each is best for. Use the default unless the user named a model or the brief clearly fits another recommendation; generating with model omitted uses it, so a plain request needs no list_models call. query searches the full live catalog for a model the user named (one page is usually enough; never page the whole catalog). model returns one model's exact settings (input schema) and credit rates. Never estimate several models to compare; if the user's model is unavailable or ambiguous, ask once.
list_voices
ChatGPTFind voices for generate_speech; not needed for a plain voiceover, which uses the default voice (Sarah) on the default model. A voice renders only on its own engine's models (compatible_models): ElevenLabs voices on Eleven models, Fish Audio voices on Fish models, Inworld voices on Inworld models. source=premade searches the managed catalog (filter by provider); source=mine lists your existing voice clones where your plan includes voice cloning (clones are made in the Voyager app).
load_skill
ChatGPTVoyager makes AI images, video, voiceover, music, sound effects and 3D. Call load_skill('voyager-cloud') before generation unless already loaded, then the matching family skill. These are the full guides, available even without a plugin installed. Use file for a referenced companion (for example references/typography.md). Available skills: voyager-cloud-3d: Cloud: generate 3D models with Voyager from a text description or a reference image, and retexture them. voyager-cloud-audio: Cloud: generate narration with premade or existing Voyager voices, music and sound effects. voyager-cloud-image: Cloud: generate or edit still images with Voyager, including branded images and images of saved Cast characters. voyager-cloud-video: Cloud: generate short AI video clips with Voyager from a prompt, a start image or references, including saved Cast characters. voyager-cloud: Cloud: Voyager account, credits, estimates, references, jobs, results, brand kits and Cast for hosted generation. Read first for any Voyager image, video, audio or 3D request in ChatGPT or Codex.
remove_cast_reference
ChatGPTRemove a reference image from a Cast card.
save_cast
ChatGPTCreate a Cast card from images you own, or replace an existing card (pass cast_id and expected_version).