Make or edit media with Pika
- Go straight to the app. Choose by the task, not the output type, and make its call first, before
get_accountor any listing:- a task in Presets below (headshot, photo effect, resize, swap, relight, color grade, product shot, camera move, talking head, extend, trim, upscale, motion transfer, World Cup, …) →
show_presets {app}with the app its line names. That shows the user the app's choices, and its preset form collects their photo or clip, so don't callshow_uploadfirst; when they already gave you a photo or clip as a URL, pass it ininputs([{slot, value}]) to prefill the form. When the user already gave everything, callcreate_from_presetwith what you have; - an ad, podcast, character, voiceover, music or multi-shot video →
pika-ad,pika-podcast,pika-character,pika-speech,pika-audioorpika-video; - anything else →
create_videoorcreate_image, which also fit when the user names a model (pass it asmodel). Calllist_appsonly when the request is unclear: it lists every app (id, name, what each makes) as data for you to suggest a few. When the user wants to look around,show_appsopens the app gallery.
- a task in Presets below (headshot, photo effect, resize, swap, relight, color grade, product shot, camera move, talking head, extend, trim, upscale, motion transfer, World Cup, …) →
- Gather inputs in one message. Map what the user already said into the tool's fields. When you already have their
aboutfromget_account, fill what they didn't say from it: the aspect ratio for where they post, their brand, their style. For anything an app tool still needs (a preset's form collects its own):- their photo, clip or audio →
upload_filefor a file attached in chat (where offered),prepare_uploadfor a file where you can run commands, otherwiseshow_upload(passkindwhen the field takes one kind,appwith the app's id, and a shortpurposesuch as "the photo becomes the first frame"); - something they uploaded to Pika before →
list_jobs {filter: "uploads"}, orshow_creations {filter: "uploads"}to let them pick; - an earlier Pika result →
show_creations; - a style, template, effect, voice or other choice the user picks →
show_presetswith the app (andcollectionwhen the field's description names one); it shows published presets and built-in choices. When you choose yourself, read them withlist_presetswith the same arguments instead, which shows the user nothing. An enum field's values are also its choices. A look the user describes works too. Open one card per turn, media first; show presets once the media is in. When they say they uploaded or picked, callget_panelwith that card'spanel_id.
- their photo, clip or audio →
- Price and run it. Say in one line what you're running, then call it with the input and a fresh
request_idand nomax_creditsand follow the quote'snext(it applies the user's approval mode). Your line sits above the card, whose Create is the user's yes: don't ask for a yes or restate the price after it. The card runs the user's Create, edits and Take it further itself; once it reports a started job, don't call the tool again — useget_job. With no card, on a yes call it again with the same input,request_idand the agreedmax_credits. The quote card shows the final prompt under "Sent to the model", where the user can still edit it. If the user is out of credits or short, followpika-credits. - Follow
next. Each result names the next call, including when to wait on the job withget_job. When a run started from the card finishes, the result and its next steps are in the card; no need to describe them. A "Take it further with …" message from the card asks you to run that app on the named result. Otherwise offer the follow-ups the result suggests.
Presets
A preset is one of Pika's curated presets: a ready-made run of an app with a few slots to fill. show_presets {app} shows the user an app's presets to pick from, and show_preset {preset_id, inputs} opens one preset's form with the slots you have filled in. When you need the data, without showing anything: list_presets {app} lists an app's presets, each with its slots: kind (image, video, audio, text or choice), whether required, how many, and the options; list_preset_options {preset_id, slot} lists a long option list; get_preset {preset_id} reads one preset with its guide, which you read before its first use. Run one with create_from_preset {preset_id, inputs: [{slot, value}], request_id}; a value is a media URL, text or an option id, and a slot repeats for several media. Price and run it as in step 3.
Pass what you already have. If required slots are missing it returns needs_input and the card's preset form collects the rest, so don't ask in text for choices the card shows. A preset needs no prompt guide; the prompt skills are for runs where you write the prompt.
Each task's app, for show_presets {app} (step 1):
- headshot from a photo →
headshot; - photo effect (squish, melt, inflate, …) →
pikaffects; - photo at a new aspect ratio or platform format →
smart-resize, one ratio per run; - replace the characters, outfits, location or products in a video →
swap-anything(acharacter_creatorsheet works as a character); - add someone or something new to a video →
pikadditions; - new framing, angle or camera move for a clip →
camera-director(seepika-camera-director); - relight a photo or clip →
relight-media(oredit_imagefor a photo); - color grade a photo or clip, or copy a reference image's grade →
color-grading; - product shot from product photos →
product-shot; - longer video →
extend-video; extend its output again for more, or plan a multi-shot video withpika-video; - copy a clip's motion onto a photo →
motion-transfer; - sharper or higher resolution →
upscale(seepika-upscale); - cut a clip →
video-trim; wait for it withget_job, then pass its output url to the next app. Cut a longer clip first for apps that take a short one (Camera Director, extend); - talking head from a face and audio →
talking-head-studio; a script is spoken withcreate_speechfirst, and its output passed once that job finishes; - World Cup fan still →
world-cup-2026; anime World Cup clip →anime-world-cup-2026.
Other jobs
- an image from text or up to 10 references at up to 4K →
create_imagewith modelseedream-4(1K,2Kor4K); - sound effect, music, or a score for a clip →
create_audio(seepika-audio); - animate between a start and an end frame →
create_videowith modelpika; - new aspect ratio for a video →
create_videowith modelseedance-25, the clip invideo_urls, the targetaspect_ratio, and a prompt to recompose the scene and extend its surroundings while keeping the subjects, action, timing and on-screen text;
To stop a running job the user no longer wants, call cancel_job. To find a job from earlier, call list_jobs; to show one, show_creations with its job_id.