Nodaro DocsUtility, recipes and skills
Test the connection with ping, load tested recipes with get_recipe, start the Film Director skill, and prepare UGC website videos with the builder tools.
The utility, recipe and skill tools help an assistant start well. ping tests the connection, get_recipe loads tested multi-step playbooks, start_film_director loads the Film Director skill, and the three UGC builders prepare a creator video about a website. None of them generates media or spends credits. Three more tools on this page do spend credits: capture_site captures a web page as screenshots, and ugc_split_speech and ugc_finish_clips cut the speech and finish the clips of a UGC video.
Other tools that return instructions are documented with their family: start_workflow_editor and get_node_skill, start_video_director, get_studio_production_skill and get_recast_authoring_skill.
ping
Returns pong, your Nodaro user id and the name of the client that called it. Use it to check that the connector works.
Permission: none, always visible. Credits: free.
This tool has no parameters.
get_recipe
Lists and loads content recipes: tested playbooks that walk an assistant through a full multi-tool flow, such as a narrated animated explainer. The tool only returns text. The actions a recipe asks for are done with the other tools, each with its own permission and price. See Content recipes for the catalog.
Permission: none, always visible. Credits: free.
| Parameter | Type | Notes |
|---|---|---|
recipe | string | A recipe name, such as video-explainer. Leave it out to list every recipe. |
file | string | A reference file inside the recipe, such as references/prompts.md. Needs recipe. |
Returns: without arguments, each recipe's name, description and trigger phrases. With recipe, the recipe's full instructions. With file, that file. An unknown name returns an error with the valid names.
On Nodaro Cloud, the list also includes a UGC website video: a creator talking to the camera about a site, with its screenshots shown as cards.
start_film_director
Returns the Film Director skill: the instructions that make the assistant direct a film in 10 stages, from the script to the final cut, and build an editable workflow on your canvas as it goes. Assistants call it first for any multi-shot video request, such as a trailer, a short film, a music video, a reel or an ad. See Film Director.
Permission: none, always visible. Credits: free. Nodaro Cloud only.
This tool has no parameters. It changes nothing.
capture_site
Captures a web page as real phone-width screenshots: one full-page image and up to 8 section stills, cut to the shape of a phone card. It also returns a map of the page's sections, with the heading, the position and a short text of each, and the exact figures from the page text. Consent pop-ups are hidden, never accepted.
Permission: workflows:execute. Credits: a flat price per capture. The tool is listed only where the install enables page capture.
| Parameter | Type | Notes |
|---|---|---|
url | string | Required. The page to capture. The https:// is optional. Up to 2,048 characters. |
max_stills | integer | The most section stills to return, from 3 to 8. Default 8. |
client_request_id | string | A retry token: 8 to 128 characters of letters, digits and _ - . :. Reuse it when you retry after a timeout, so the capture is not started or charged twice. |
Returns: a job id. Read the stills from outputData.stills of wait_for_job, each with an assetId and a url. The stills are saved in your media, hidden from the picker until you save them, and they count toward your storage.
A page that blocks automated visitors, or that comes back empty, fails and is refunded. A page whose robots.txt does not allow the capture is refused before anything runs.
The UGC video builders
Three free helpers used by the UGC website recipe. They generate nothing: each one checks or prepares its part and returns the exact arguments of the next generation calls. Two more tools, ugc_split_speech and ugc_finish_clips, do the audio and video work of the same recipe for a flat price per call. All five are available on Nodaro Cloud only, and need no permission, except build_ugc_creator with a saved character, which needs assets:read.
build_ugc_creator
Prepares the creator who talks to the camera: a newly sampled person, one of your saved characters, or your own photo.
| Parameter | Type | Notes |
|---|---|---|
source | string | Required. sampled, character or photo. |
character_id | string | For character: the saved character. |
photo_url | string | For photo: your photo. |
gender | string | woman or man. |
product_category, overrides, previous, seed | Optional details to steer or repeat the sampled creator. |
Returns: the creator's identity images, or the exact arguments of the image generations to run next. For a sampled creator, that is one image per candidate, so the user can choose one.
build_ugc_clips
(Cloud only) Checks a UGC video script — every problem at once, each with a fix — and returns one generate_video call per clip, the call that joins them, and a credit quote for the rest of the video. Scripts can be written in English or Hebrew. A Hebrew script is answered in two steps: the first call returns the speech to record and how to split it, with the quote; call again with the segments from ugc_split_speech to get one generate_video call per segment. Ungated, free. Used by the ugc-website recipe (get_recipe).
| Parameter | Type | Notes |
|---|---|---|
plan | object | Required. The script. |
gender | string | Required. woman or man. |
identity_images | array | Required. One or two images of the creator. |
traits, seed | Optional. | |
spent_job_ids | array | The jobs already run for this video, which the quote lists as spent. Up to 64. |
segments | array | The segments from the output of ugc_split_speech, when the answer to the first call asked for them. Leave it out on the first call. Up to 8. |
speech_lane | string | English only. reference-audio (default): a voice speaks the script, and the video model performs it. prompt-quoted: the video model reads the line in its own voice. |
voice_id | string | English only. The voice of the reference-audio lane, a premade name or an id from list_voices, instead of the voice the builder picks for the creator. |
Returns: one ready-to-run generate_video call per clip, the call that joins the clips, and a credit quote for the rest of the video. A quote line with estimate: true is the estimate of a metered LLM call, the screenshot reading's Image to Text call. That call is charged for the tokens it actually used. The quote then carries totalIsEstimate: true, so show the total as ≈N. A model with no price is refused with llm_model_unpriced.
build_ugc_cards
Times the screenshot cards and the captions of a UGC video from its word timings.
| Parameter | Type | Notes |
|---|---|---|
plan | object | Required. The script. |
alignment or words | array | The word timings, from forced_alignment or from transcribe. |
video_duration_ms | number | The length of the video. |
caption_style | string | How the captions after the opening line look: body-captions (word by word, the default), karaoke, clean-subtitles or none. The opening line always keeps its plate. |
Returns: the layers for overlay_images and the caption segments for add_captions.
ugc_split_speech
Cuts the recorded speech of a UGC video into segments at its own pauses. Each segment is saved as an audio file, with a short silence on each side.
Credits: a flat price per call.
| Parameter | Type | Notes |
|---|---|---|
plan | object | Required. The script. |
audio_url or audio_asset_id | string | The speech, as a URL or as a Nodaro job or upload id. |
alignment | array | Required. The word timings of the speech, 1 to 2,000 items of { word, start, end }. |
Returns: a job whose output lists the segments, for the segments parameter of build_ugc_clips.
ugc_finish_clips
Checks each rendered clip against the audio it was given, puts that original audio on the clip, trims the dead air between words and returns the finished clips in order. If a clip does not match its audio, the tool names the clips to render again and changes nothing.
Credits: a flat price per call.
| Parameter | Type | Notes |
|---|---|---|
segments | array | Required. The segments from ugc_split_speech, 1 to 8. |
clips | array | Required. 1 to 8 items. Each has clip (its number, from 1 to 8) and the rendered video as video_url or video_asset_id. reroll_video_url or reroll_video_asset_id is a clip rendered again. |
On a Hebrew video, which is made of segments, the tool also gives each clip the sound of its setting: a reverb matched to the setting, a level of -16 LUFS and room tone underneath. Outdoor, car and hall settings get the reverb and the level, but no room tone. If this step fails, the clips are returned without it, with a warning.
Returns: a job whose output lists the finished clips in order.
Frequently asked questions
How do I check that my assistant is connected to Nodaro?
Ask it to call ping. The tool answers with pong, your Nodaro user id and the name of your client, and it needs no permission.
What does get_recipe return?
Instructions, not media. Without an argument it lists the recipes with their descriptions and trigger phrases; with a recipe name it returns that recipe's full playbook, which the assistant then follows with the other tools.
Do the UGC builder tools generate the video?
No. The three builders are free helpers of the UGC website recipe. They prepare the creator, check the script and time the cards and captions, and they return the exact calls to make next, with a credit quote. Two other tools, ugc_split_speech and ugc_finish_clips, cut the speech and finish the clips for a flat price per call.
Related
Content recipes
Film Director
MCP tools reference
Troubleshooting
Last updated on
3D scenes
Parameters of the four 3D scene tools that author an editable clay scene from a prompt, revise it, render it to MP4, or produce a finished shot in one job.
MCP conversations
Each conversation an AI assistant has with Nodaro over MCP becomes a workflow of connected nodes that you can open in the editor, change and run again.