APIhub Video

Use this when the user wants captions (subtitles) burned into a video of theirs — "add Hebrew captions", "make it a reel with subtitles". One render: the words are taken from the speech. Ask first, unless the user said: the caption style (offer classic for calm talk, bold for energetic, karaoke for words that light up; to show the styles, call show_examples with kind caption_styles — never a text list), the spoken language when unclear, and the shape (vertical 9:16 for Reels/TikTok by default, or the original). Costs credits: first call get_price with the same arguments, tell the user the price and wait for their OK, then call this with that answer's estimate_id as approved_estimate_id and approved_credits at least its credits_to_reserve — without them nothing starts. A limit the user named earlier ("up to 20 credits", "don't ask me again") covers only a price within it: above it, do not start — tell them the real price and ask. Answers at once with the job; the video takes a few minutes. A video the user attached in this chat: pass it as video_file (no upload link needed). Irreversible: it holds up to the approved credits and charges what the job used; spent credits, those of finished steps included, are not given back. It reaches outside the account: it reads the user's video from their link (any website) or their upload, and outside AI services transcribe its speech. It makes a new video file whose link works for 24 hours; it posts and sends nothing and overwrites or deletes none of the user's files. The same request with the same approved_estimate_id returns the same job, never a second charge.

captions_for_video

Use this to follow a video job that captions_for_video, reel_from_long_video, video_from_idea or full_edit_video started (the id in their answer): still being made (with progress and minutes left), done (with the download link), or failed (a failed job is not charged). Check every 10–20 seconds while it runs; the user may also close the chat and ask later. It reports where a job of this account stands (being planned or made, waiting for an OK or for credits, done with its link, or failed). It is not read-only only because, for a job that has just finished, it records that job's result in the user's account: its video link and its final charge, which the job the user approved had already incurred (APIhub's render servers record the same on their own). It spends nothing new and starts, changes, cancels or deletes nothing; the same check again gives the same answer; the only outside service it asks is the one making that job.

video_status

Use this to check an upload link from start_upload: waiting (no file yet), uploading, uploaded (then pass the code as `video`), or expired. Read-only: it reads one upload of this account; it changes nothing and reaches nothing outside the account.

upload_status

Use this when the user asks how many credits they have: the balance of their APIhub account in credits — available (what new jobs can use), reserved (held by jobs still running) and total. Free. Read-only: it reads this account's balance; it starts and charges nothing (it may first finish the bookkeeping of a job that already ended) and reaches nothing outside the account.

get_balance

Use this when the user wants their video fully edited — "a beautiful edit", "with AI shots / effects / music", "make mine like this example": APIhub's video agent watches the video and makes ONE finished video with AI-made shots (or stock footage) of what is said, music with sound effects, a headline in the first seconds, captions, effects like zoom punch-ins and flash cuts, and the pauses and retakes cut. One approval covers all of it: watching, planning and making. Ask only for the video and, if the user did not say, nothing else — the defaults are the full edit; pass their wishes as request, a show_examples card's id as example, and length_seconds or shape only when they named them. Costs credits: first call get_price with the same arguments, tell the user the price and wait for their OK, then call this with that answer's estimate_id as approved_estimate_id and approved_credits at least its credits_to_reserve — without them nothing starts. A limit the user named earlier ("up to 20 credits", "don't ask me again") covers only a price within it: above it, do not start — tell them the real price and ask. Then follow it with video_status (or job_status, as its answer says): planning takes about a minute, the video a few minutes more; it never posts anything. A video the user attached in this chat: pass it as video_file (no upload link needed). Irreversible: it holds up to the approved credits and charges what the job used; spent credits, those of finished steps included, are not given back. It reaches outside the account: it reads the user's video from their link (any website) or their upload, and outside AI models and stock libraries supply its shots and music. It makes a new video file whose link works for 24 hours; it posts and sends nothing and overwrites or deletes none of the user's files. The same request with the same approved_estimate_id returns the same job, never a second charge.

full_edit_video

Use this only when the user cannot attach the video in this chat — attaching failed, or the file is very large — or for a photo: a video attached here goes straight to the task tool as video_file. Returns a short link to a page where they pick the file; it goes to their APIhub account. Give them the link, wait until they say it is done (or check upload_status), then pass the code as `video` to the task tool. The link works for 2 hours. Free. It makes a new upload link in the user's own APIhub account each time (so it is not idempotent); the file picked there goes only to that account. It changes, sends and deletes nothing else, and charges nothing.

start_upload

Use this when the user wants a short faceless video from an idea and has no footage: a voice reads a script, stock footage fills the screen line by line, captions follow the words, vertical 9:16. Ask first: the idea, the video's language, and a woman's or a man's voice. Then write a script of about 60–150 words, show it to them and use the version they approve. Costs credits: first call get_price with the same arguments, tell the user the price and wait for their OK, then call this with that answer's estimate_id as approved_estimate_id and approved_credits at least its credits_to_reserve — without them nothing starts. A limit the user named earlier ("up to 20 credits", "don't ask me again") covers only a price within it: above it, do not start — tell them the real price and ask. Making it takes a few minutes. Irreversible: it holds up to the approved credits and charges what the job used; spent credits, those of finished steps included, are not given back. It reaches outside the account: an outside voice service reads the script, and stock-footage libraries are searched for its shots. It makes a new video file whose link works for 24 hours; it posts and sends nothing and overwrites or deletes none of the user's files. The same request with the same approved_estimate_id returns the same job, never a second charge.

video_from_idea

Use this when the user wants short vertical reels (Reels / TikTok / Shorts) from a longer video of theirs — a podcast, a talk, a webinar, a live. Makes 1–3 reels of 15–90 seconds, each with captions and a headline; APIhub watches the video and picks the strongest self-contained moments, unless the user names the parts (moments). A clip of 90 seconds or less becomes one reel as it is. Ask first: how many reels, about how long, the caption style, and whether they want particular moments. Costs credits: first call get_price with the same arguments, tell the user the price and wait for their OK, then call this with that answer's estimate_id as approved_estimate_id and approved_credits at least its credits_to_reserve — without them nothing starts. A limit the user named earlier ("up to 20 credits", "don't ask me again") covers only a price within it: above it, do not start — tell them the real price and ask. The reels take a few minutes. A video the user attached in this chat: pass it as video_file (no upload link needed). Irreversible: it holds up to the approved credits and charges what the job used; spent credits, those of finished steps included, are not given back. It reaches outside the account: it reads the user's video from their link (any website) or their upload, and outside AI services transcribe it and find its strongest moments. It makes a new video file whose link works for 24 hours; it posts and sends nothing and overwrites or deletes none of the user's files. The same request with the same approved_estimate_id returns the same job, never a second charge.

reel_from_long_video

Use this before captions_for_video, reel_from_long_video, video_from_idea or full_edit_video: the price of that job in credits for exactly the arguments you will pass it — free; nothing runs and nothing is charged. Name the tool in `task` and give its arguments. A video the user attached in this chat goes in video_file: it is stored in their APIhub account once, and the answer names it by an upload code (in `call`) to use from then on. Returns the price (estimated_credits, and credits_to_reserve held while it runs), the account's balance and whether it is enough, and an estimate_id the tool needs once the user says yes. It prices one job, starts nothing and charges nothing. It is not read-only only because a video file sent from this chat is saved to the user's APIhub account as an upload (sent again within two hours, the same file is the same upload). It opens the user's video link, which can be on any website, to measure the video.

get_price

Use this when the user asks for examples, what APIhub can make, which styles there are ("show me styles", "which captions can I use"), or is unsure what to make — never answer those with a text list or a table of names. kind best (default): our best finished videos, each with what it cost in credits. kind caption_styles: every caption style as a running preview, to pick one. Free; nothing runs. Each card says what this chat makes from it (full_edit_video in its look, or captions_for_video with its style). Where the app shows it, the videos play in a row with a button that asks for one like it; otherwise give the user the links. Read-only: it shows APIhub's own example videos and caption styles; it changes nothing and reaches no outside website.

show_examples