API reference

Authentication

Every request needs a personal access token in the Authorization header. Create one in your dashboard under Settings → Connect & API. The token starts with cf_live_ and is shown once, at creation.

Authorization: Bearer cf_live_... 

A token has full access to your account. Keep it server-side, never in a browser or a public repo. Revoke it from the same page if it leaks. The same token also works for the MCP server.

Quickstart

From idea to rendered video in four calls. First, check what your account can do:

curl https://api.channel.farm/v1/account \
  -H "Authorization: Bearer $CF_TOKEN" 

curl -X POST https://api.channel.farm/v1/scripts \
  -H "Authorization: Bearer $CF_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"idea": "The hidden history of coffee", "target_length_minutes": 5}' 

curl https://api.channel.farm/v1/scripts/<script_id> \
  -H "Authorization: Bearer $CF_TOKEN" 

curl -X POST https://api.channel.farm/v1/videos \
  -H "Authorization: Bearer $CF_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"brand_profile_id": "<id>", "script_id": "<script_id>", "pacing": "dynamic"}' 

You need at least one brand profile with a voice set before rendering. Create one with POST /brand-profiles or in the dashboard.

Errors

Errors are JSON with a stable code and a human-readable message. When subscribing or upgrading clears the error, the body also carries pricing_url.

{
  "error": {
    "code": "credit_quota_exhausted",
    "message": "You've used all 750 credits in your Starter plan for this billing period. ...",
    "pricing_url": "https://channel.farm/pricing",
    "action_required": "subscribe_or_upgrade"
  }
} 

Credits & rate limits

Rendering spends credits, the same currency as the dashboard. A render costs about minutes × rate for its pacing, never less than the minimum: 25 credits for a 16:9 video and 8 for a 9:16 one (GET /account has the live figures):

Animated scenes cost extra on top of that. Each animated scene is charged by its length, so turning animation on can multiply the price of a long video. Send your settings to POST /scripts/validate to see the real figure before you spend anything.

Check your balance with GET /credits and preview a script's cost with POST /scripts/validate. Script generation is capped at 30 per hour and renders at 10 per hour. Reads are not metered. Rendering requires an active paid plan. Without one, reads and profile management still work.

Polling

Script generation and rendering are asynchronous. The create call returns immediately with an id. For videos, POST /videos also returns estimated_build_seconds and recommended_first_poll_seconds. Wait that long before the first poll, then follow next_poll_after_seconds from each GET /videos/:id response. Renders typically take 5 to 40 minutes depending on length and load.

Estimates use recent comparable generations and exclude time spent editing. With advanced editing enabled, the estimate ends at ready_to_edit; otherwise it ends when the video is rendered. Estimate fields may be null when timing data is unavailable; the polling recommendation still applies.

Plan, credits, quotas, and limits.

{
  "subscription_tier": "starter",
  "subscription_status": "active",
  "can_render_video": true,
  "plan": { "max_video_length_min": 15, "videos_per_month": 6 },
  "credits": {
    "allowance": 750,
    "used": 210,
    "remaining": 540,
    "rollover": 0,
    "period_end": "2026-09-01T00:00:00+00:00",
    "bonus": { "balance": 0, "soonest_expiry": null, "purchased": 0, "gifted": 0 },
    "rates": {
      "per_minute": { "relaxed": 5, "dynamic": 7, "rapid": 9, "hyper": 12 },
      "minimum_per_video": 25,
      "minimum_per_short_video": 8,
      "note": "A render costs about minutes × rate for its pacing, never less than the minimum..."
    }
  },
  "script_quota": { "used": 3, "limit": 60, "remaining": 57, "unlimited": false },
  "input_limits": { "idea_min_chars": 5, "idea_max_chars": 2000, "script_max_video_length_min": 15 }
} 

Credit balance for this billing period.

The credits block on its own, for cheap balance checks between renders. allowance includes any rollover. bonus is gifted or topped-up credits, spent after the plan allowance. Null allowance means no active billing period.

{
  "subscription_tier": "starter",
  "can_render_video": true,
  "allowance": 750,
  "used": 210,
  "remaining": 540,
  "rollover": 0,
  "period_end": "2026-09-01T00:00:00+00:00",
  "bonus": { "balance": 100, "soonest_expiry": "2026-08-20T00:00:00+00:00", "purchased": 0, "gifted": 100 },
  "rates": {
    "per_minute": { "relaxed": 5, "dynamic": 7, "rapid": 9, "hyper": 12 },
    "minimum_per_video": 25,
    "minimum_per_short_video": 8,
    "note": "A render costs about minutes × rate for its pacing, never less than the minimum..."
  }
} 

List brand profiles.

A brand profile is the reusable identity videos are rendered with: voice, visual style, captions, and optionally a cast of recurring characters. render_ready is true only when the profile has a voice that can narrate: an ElevenLabs voice, a Google voice, or one of your own. voice_tags says which delivery marks that voice performs (see GET /brand-profiles/:id/voice-tags).

{
  "brand_profiles": [
    {
      "id": "9a1e...",
      "name": "Sleepless Histories",
      "is_favorite": true,
      "language": "en",
      "render_ready": true,
      "voice": { "id": "41b0...", "name": "Archer", "provider": "elevenlabs", "custom": false },
      "voice_tags": "eleven-v4",
      "visual_style": { "id": "77aa...", "name": "Storybook Ink" },
      "character": {
        "id": "c3f2...",
        "name": "Professor Owl",
        "status": "approved",
        "style_matches_profile": true
      }
    }
  ]
} 

Create a brand profile.

Counts against your plan's brand-profile limit. A new account gets its first profile free. Returns 201 with the new brand_profile_id.

Update a brand profile.

An ElevenLabs voice at speed 1 renders on Eleven v4 Turbo, which uses stability and similarity_boost and performs voice tags. Any other speed moves it to multilingual-v2, which uses all four dials but would read a tag out loud, so it takes pauses only. Hear a setting before saving it with POST /voices/preview.

curl -X PATCH https://api.channel.farm/v1/brand-profiles/<id> \
  -H "Authorization: Bearer $CF_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"caption_size": 48, "voice_tuning": {"stability": 0.4}}' 

Every setting of one brand profile.

Every setting of one profile, in the same terms PATCH takes: the voice and the model it renders on, its tuning or delivery style, which voice tags it performs, the visual style, the captions for 16:9 and 9:16, and the cast. Read it before you change anything.

{
  "brand_profile_id": "9a1e...",
  "name": "Sleepless Histories",
  "language": "en",
  "render_ready": true,
  "voice": { "id": "41b0...", "name": "Archer", "provider": "elevenlabs", "custom": false },
  "voice_tuning": { "stability": 0.5, "similarity_boost": 0.75 },
  "voice_style": null,
  "voice_model": "Eleven v4 Turbo (stability and similarity_boost apply; voice tags work)",
  "voice_tags": "eleven-v4",
  "visual_style": { "id": "77aa...", "name": "Storybook Ink", "is_yours": false },
  "captions": { "enabled": true, "font": "Inter", "size": 35, "color": "#ffffff", "highlight_color": "#00ff00", "shadow": "soft", "words_per_line": 4, "uppercase": false,
    "position": { "x_pct": 50, "y_pct": 23.9, "max_width_pct": 96, "quick_position": "top-center" } },
  "short_captions": null,
  "cast": [
    { "id": "c3f2...", "name": "Professor Owl", "enabled": true, "lead": true, "mention": "@professor-owl", "style_matches_profile": true }
  ]
} 

Delete a brand profile for good.

Permanent. It also deletes the characters that were created on this profile and the scripts written for it, and the response counts them. Videos already made are kept. Refused with a 409 while one of its videos is still being made.

Set a profile’s cast of characters.

The whole cast of recurring characters, replacing what was there. The first one switched on is the lead. Every new video from the profile uses the switched-on cast, and character_ids on POST /videos can pick a different cast for one video. Each must be your own approved character. One that is switched on must be drawn in the profile’s visual style; one switched off can stay on the list.

The voice tags a profile’s voice performs.

Which delivery marks this profile’s voice performs, with every one of them and a hint. Write them into the script you render. They are performed, never spoken, and never shown in the captions.

A set of null means this voice is offered no marks right now: write plain narration. POST /scripts/voice-tags writes them in for you.

List recurring characters.

A recurring character looks the same in every video it appears in. Draw one with POST /characters, save it with POST /characters/:id/approve, and set which characters a profile uses with POST /brand-profiles/:id/characters. This list shows each character and the profile casts it is in.

{
  "characters": [
    {
      "id": "c3f2...",
      "name": "Professor Owl",
      "status": "approved",
      "attachable": true,
      "mention": "@professor-owl",
      "preview_image_url": "https://...",
      "visual_style_id": "77aa...",
      "visual_style_name": "Storybook Ink",
      "created_for_brand_profile_id": "b1d0...",
      "in_casts": [{ "brand_profile_id": "b1d0...", "brand_profile_name": "History Owl", "enabled": true, "lead": true }],
      "created_at": "2026-07-25T18:12:00+00:00"
    }
  ]
} 

Browse narration voices.

About 1,200 ElevenLabs voices, over 2,000 Google voices in many languages, and your own voices (is_yours). Featured voices come first. A voice with a null preview_url gets one from POST /voices/preview. Use the id as voice_id.

Hear a voice.

Returns an audio_url: the same sample line, in the voice’s own language, on the model your videos render with. With no dials or style to try, a library voice returns its standard sample. Free. Limited to a few a minute and a daily number.

Design or clone a voice of your own.

A Google voice only your account can use. Starter plan and up. Each voice costs 15 credits, taken when it starts and refunded if it fails. Each plan includes a number of voices: 1 on Starter, 2 on Developmental, 3 on Professional. The response has a sample_url to listen to.

I am the owner of this voice and I consent to Google using this voice to create a synthetic voice model. 

Delete one of your own voices.

Deletes one of your own voices for good, with its recordings. Refused with a 409 while a brand profile still uses it: change the voice on those profiles first. The credits it cost are not returned. Videos already made keep their narration.

List visual styles.

The stock style catalog plus your own custom styles (is_yours). Each has a sample image_url showing the look. Filter with ?category=.

Create a custom visual style.

Custom styles are private to your account, and a paid feature. Limit 20. The pictures steer the look of every scene: the example pictures when you send them, and otherwise the card art. Three example scenes are drawn in the background after it is saved.

Change one of your own styles.

Only your own styles. When the look changes on a style without example pictures, its card is drawn again and new example scenes follow. Videos already made keep their look. Example pictures cannot be changed: make a new style for a new set.

Delete one of your own styles.

Deletes one of your own styles for good. A style that has already made videos cannot be deleted (409). Brand profiles still using it are named in the response and need another style before they can make new videos.

Generate a script from an idea.

Returns 202 with a script_id. Poll GET /scripts/:id until ready is true. The defaults match the create page’s Write button. Counts against your script quota. Capped at 30 per hour.

Fetch a generated script.

The script comes back with any voice tags the writer added. A script that ran past your plan’s length is cut at a whole sentence, the way the create page cuts it, and rendering this script_id renders exactly that text.

{
  "script_id": "5b7c...",
  "status": "completed",
  "ready": true,
  "script": "Every morning, two billion people...",
  "title": "The Hidden History of Coffee",
  "topic": "The hidden history of coffee",
  "language": "en",
  "estimated_seconds": 482,
  "has_voice_tags": false
} 

Estimate length and credit cost.

Estimate a script's spoken length, whether it fits your plan, and what it would cost to render at each pacing. Free to call. It measures the words that are spoken: voice tags and cues are not counted. Send language or the brand_profile_id you will render with, because languages are spoken at different speeds.

You can also send any of the settings from POST /videos. You then get back estimated_credits_for_settings: the price of that exact video, animation included. Do this before you animate anything.

With brand_profile_id, it also checks the script’s tags against that profile’s voice. Tags written for a different voice come back in foreign_voice_tags; POST /videos refuses the script until they are fixed. Send fix_foreign_tags here to see and price the fixed script first.

{
  "language": "en",
  "estimated_seconds": 292,
  "estimated_minutes": 4.9,
  "plan_max_minutes": 15,
  "fits_plan": true,
  "estimated_credits": { "relaxed": 25, "dynamic": 34, "rapid": 44, "hyper": 59 },
  "estimated_credits_for_settings": 96,
  "delivery_marks": { "tags": 3, "cues": 0 }
} 

List your saved scripts.

Your saved scripts, newest first. Any one with ready: true renders with script_id on POST /videos.

Delete a saved script.

Deletes a saved script. Videos already made from it are unaffected, and it still counts toward this month’s script allowance. A script that is still being written cannot be deleted yet (409).

Have AI add voice tags or cues to a script.

The create page’s “Add with AI”. An AI adds [tags] for an ElevenLabs voice or <cues> for a Google voice, and never changes a word. Free, within a daily number of runs. Show the result to the person, then send it as script_text to POST /videos.

Draft a visual direction.

The create page’s “Ask AI for a draft”: the world the scenes are drawn in, written from the script. Free, needs a paid plan, and has an hourly limit. It takes 20 to 60 seconds, so poll GET /visual-directions/:id. Edit it with the person, then send it as visual_direction on POST /videos.

Poll a visual direction draft.

{
  "visual_direction_id": "0c1d...",
  "status": "completed",
  "ready": true,
  "direction": "A rain-soaked harbour town in the 1920s...",
  "error": null
} 

Render a video.

{
  "video_id": "e8d4...",
  "status": "processing",
  "title": "Coffee history",
  "pacing": "dynamic",
  "language": "en",
  "settings": {
    "pacing": "dynamic",
    "animate": "first",
    "animate_first_scenes": 3,
    "transition": "fade",
    "motion": "zoomIn"
  },
  "edit_before_render": false,
  "estimated_credit_cost": 95,
  "estimated_build_seconds": 1348,
  "estimated_build_minutes": 22.5,
  "recommended_first_poll_seconds": 1348,
  "recommended_poll_interval_seconds": 30
} 

This is the one paid action: it reserves the credit cost up front. Requires an active paid plan or bonus credits. Capped at 10 renders per hour. A script longer than your plan allows is refused with a 402; a generated script is first cut to fit at a whole sentence, as the create page does.

Everything except brand_profile_id and the script is optional. Send nothing else and you get the same video this endpoint always made. The settings field in the response is what the render actually started with, defaults filled in.

With edit_before_render set, the video stops at ready_to_edit and waits. Nothing finishes it until you export it, so only set it if you plan to edit.

List recent videos.

Most recent first.

Poll a video render.

{
  "video_id": "e8d4...",
  "status": "completed",
  "ready": true,
  "progress_percent": 100,
  "next_poll_after_seconds": null,
  "video_url": "https://storage.googleapis.com/.../final.mp4",
  "youtube_url": null,
  "thumbnail_url": "https://...",
  "duration_seconds": 294,
  "youtube_metadata": {
    "title": "The Hidden History of Coffee",
    "description": "...",
    "tags": ["coffee", "history"]
  },
  "error": null
} 

List a video’s scenes.

Every scene of a finished video, in order: the prompt it was drawn from, a link to the image, the narration spoken over it, how long it lasts, and whether it has been animated.

Scenes are numbered from 0. That number is how you refer to a scene everywhere else, so you do not have to keep track of ids. A motion of null means the scene follows the video setting, which is not the same as "none".

{
  "video_id": "e8d4...",
  "status": "ready_to_edit",
  "editable": true,
  "scene_count": 34,
  "scenes": [
    {
      "scene": 0,
      "image_id": "9f2a...",
      "prompt": "A hand pouring coffee...",
      "narration": "Every morning, two billion people...",
      "image_url": "https://storage.googleapis.com/.../image_0.png",
      "seconds": 14.2,
      "motion": null,
      "transition_after": null,
      "animated": false,
      "ready": true
    }
  ]
} 

List a video’s narration lines.

The narration, line by line. Only videos with segmented narration have separate lines. An older video returns segmented: false and its narration cannot be changed.

Line numbers have gaps where lines were deleted. Read seq from this response rather than counting.

Each line says which voice tags its voice reads in tag_family: elevenlabs ([tags]), google (<cues>) or none. A line that carries tags also has spoken_text: the line as its voice is given it.

Open a finished video for editing.

Open a finished video so it can be changed. Nothing can be edited until you do.

The video moves to ready_to_edit and stays there. The version people can watch is untouched until you export, and nothing else will finish it. Safe to call twice. A video created with edit_before_render is already open.

Render your edits.

Render your edits and make them the version people see. This is the only thing that finishes a video waiting at ready_to_edit. Poll GET /videos/:id until it is completed.

A video can only be exported a limited number of times, and the response says how many are left. A 409 with voiceover_master_stale means a narration edit is still settling: wait a few seconds and call it again.

Draw a scene again.

On a black screen this fills it with a new picture from prompt. Write @name in a prompt to put one of the video’s characters in.

Each video gets a number of free scene drawings, 10 by default, shared with POST /videos/:id/scenes/add, and does not get more. After those, a drawing costs credits and goes ahead rather than being refused; the response says which happened. Going back to an earlier picture is free, so check the versions first. The new picture takes a minute. Poll GET /videos/:id/scenes until that scene is ready.

A scene’s earlier pictures.

Every picture drawn for one scene, oldest first, with the prompt each came from and which one is showing. Going back to one is free, so check here before spending another redraw.

Go back to an earlier picture.

Show an earlier picture instead of the current one. Send the file_path from the versions list. Free, instant, and reversible.

Cut a scene into two.

Cut one scene into two showing the same picture, so a long stretch of narration can have two pictures. The video does not get longer: the scene’s time is divided between the halves. Send at_seconds to choose where, measured from the scene’s own start; it defaults to halfway.

Both halves show the same picture until you regenerate one, which gives it a picture of its own automatically. Scene numbers after the split shift up by one.

Take a scene out.

Take a scene out. Its time goes to the scene before it, so the video stays exactly as long and the narration is untouched: the previous picture just stays on screen longer. Scene numbers after it shift down by one.

To hide a scene without giving its time away, black it out instead. To make the video shorter, delete the narration.

Black out a scene.

Replace a scene’s picture with a black screen for exactly as long as that scene lasted. Nothing else moves and the narration keeps playing. Fill it again later with a new drawing or your own image.

One scene’s motion, transition, fades.

Setting a transition clears any fade at that cut, because a fade would otherwise outrank it and the transition would never play.

The whole video’s cuts and motion.

Works like the editor’s pickers. A new transition applies to every cut and clears the per-cut ones you had pinned. A new motion applies to every still scene, clears the per-scene ones, and keeps the current speed unless you send one.

Turn a still into a moving clip.

Turn one scene’s still picture into a short moving clip. This spends credits. They are charged when it starts and are not refunded if you change your mind, and the amount scales with how long the scene runs.

Send quote_only: true first to see the price without spending anything. Needs a Starter plan or above. It takes a minute or two; poll the scenes list until that scene shows animated: true.

Back to the still picture.

Put an animated scene back to its still picture. The credits already spent are not refunded, so this undoes the look, not the cost.

Draw a new picture into the video.

The scene before it is cut in half, so the video does not get longer and the narration is untouched. Use it when a stretch of narration needs a second picture, not to make the video say more. Scene numbers after the new one shift up by one.

Each video gets 10 free scene drawings, shared with POST /videos/:id/scenes/regenerate. After those, every drawing costs credits. The response says which happened, in credits_charged and free_drawings_remaining. The picture takes a minute; poll GET /videos/:id/scenes until it is ready.

Put your own image in a scene.

Put your own photo or artwork into a scene, or into a black screen, for the times no prompt would produce it. JPG, PNG or WebP, at least 320 by 180, up to 10 MB. Free, and the picture it replaces is kept, so POST /videos/:id/scenes/versions can put it back.

Capped at 30 uploads an hour across your account, and 200 per video.

Move the cut between two scenes.

The time moves between the two scenes, so the video’s length and its narration never change. Use it when a picture arrives too early or lingers.

If the amount you asked for would push a neighbour under its minimum length, it is trimmed to fit. The response then has clamped: true and moved_seconds is what actually happened.

Move a scene somewhere else.

Works like dragging a scene in the editor. The cuts stay where they are on the narration and black screens stay put, so the picture you moved fills its new slot’s length and plays over different words. Re-read GET /videos/:id/scenes afterwards and check each picture still matches what is being said.

Say a line again, or differently.

This is the only way to change what the video says. If the new take is a different length, the video’s length changes with it and the pictures shift to match, so keep new words close to the original length. It takes a few seconds. Poll GET /videos/:id/narration until the line is ready, then re-read the scenes before judging the timing.

Add a spoken line.

The video gets longer by however long the line takes to say. No scene is added: the picture covering that moment simply stretches over the new words, so follow with POST /videos/:id/scenes/add if they deserve their own picture. Line numbers after it shift up by one.

Cut a spoken line out.

The video gets shorter by exactly that line and the pictures close up over the gap. This is how you cut something the video should not say. Removing a scene does not shorten anything.

It is free and does not use up a narration edit. The line numbers left behind keep their gaps.

Silence a line, keep its pause.

The words stop, the pause stays, and nothing else moves. Use it to drop a sentence while leaving every picture exactly where it is. To take the time back as well, delete the line instead.

Title, description, tags, thumbnail.

GET /videos/:id says whether a video is done. This says whether it is any good: the title, description, tags and chapters it will publish as, the thumbnail and the prompt behind it, where it is on YouTube, and how many regenerations of each are left.

{
  "video_id": "e8d4...",
  "status": "completed",
  "title": "The Hidden History of Coffee",
  "description": "...",
  "tags": ["coffee", "history"],
  "thumbnail_url": "https://...",
  "thumbnail_prompt": "THE PRESENTER holds up a steaming coffee cup...",
  "thumbnail_text": "NOBODY TOLD YOU",
  "chapters": [{ "title": "Intro", "start_seconds": 0 }],
  "youtube_url": "https://www.youtube.com/watch?v=...",
  "archived": false,
  "marked_published": true,
  "regenerations_left": { "title": 2, "description": 3, "tags": 3, "thumbnail": 1 }
} 

Write the title, description or tags again.

Free, but capped at 3 regenerations of that field per video, for ever. It takes a few seconds. Poll GET /videos/:id/metadata until the field changes. The old value is kept, so a regeneration is always reversible.

Draw the thumbnail again.

The thumbnail is drawn in the video’s own style, from its first scenes, with its recurring character as the constant. Anything you leave out keeps what the video already has. Free, but capped at 3 thumbnail regenerations per video, for ever. Vertical 9:16 videos keep their first scene and are refused. It takes about a minute. Poll GET /videos/:id/metadata until thumbnail_url changes.

Earlier titles, tags and thumbnails.

Every title, description, tag set and thumbnail this video has had, oldest first, with is_current marking the one showing now. Always check here before spending another regeneration.

Put an earlier one back.

Free, instant, and it spends no regeneration. The value it replaces is kept too, so this is reversible in both directions.

Rename a video.

Your own label for the video, the name the library and GET /videos list it under. It is never sent to YouTube: the YouTube title, description and uploads stay as they are. A title sent to POST /videos is kept as this name.

Hide a video, or bring it back.

Reversible, and it destroys nothing: the video, its file and its history all survive. This is the one to reach for whenever a video just needs to be out of the way.

Mark a video as published.

For a video you posted somewhere yourself: it only marks the video in your library. It posts nothing anywhere. To put a video on YouTube, use POST /youtube/posts.

One video’s captions.

The captions of this one video, like the editor’s Captions panel. The brand profile is not changed. The video must be open for editing, and the new captions show once you export it. The response’s captions.position is where they sit ({ x_pct, y_pct, max_width_pct }), or null for the normal spot at the bottom centre. The box width is worked out for you, so the text stays on screen.

Delete a video for good.

Permanent, and it takes the scenes, the narration and the edit history with it. This cannot be undone, and the credits the video cost are not returned. Use POST /videos/:id/archive unless the video is genuinely meant to be destroyed.

A video that is still building cannot be deleted. Wait for it to finish or fail first: the credits it reserved are spent either way.

Draw a recurring character.

Each drawing comes out of a monthly free allowance, then costs a credit, and the response shows both, so settle the description first. Calling it again adds another take of the same character rather than making a second one, so this is both "make one" and "try again". The response says how many redraws are left. Needs an active paid plan.

The description it returns is not the one you sent: it is what we read back off the picture, and it is the text that travels into every scene. Judge that as much as the image. Nothing renders in a video until you approve a take.

Reword a character, free.

Free and instant. Nothing is drawn and nothing is charged. Use it to settle the wording before spending a draw, because the description is what travels into every scene prompt. Pass the result to POST /characters, or to POST /characters/:id/approve to save it against a picture already drawn.

Save a character.

Saving fixes the picture and the description for good. Only the name can be changed afterwards, because a saved character has to stay the one the person judged. Be sure here, not after.

Approving also makes it the lead character of the profile it was drawn on, so every new video from that profile shows it. To use it in another profile, add it to that profile’s cast with POST /brand-profiles/:id/characters. A character only renders in the visual style it was drawn in.

Throw a character away.

Throws a character away, saved or not. Any brand profile using it stops using it, so its next videos are drawn without a recurring character. Videos already made keep theirs.

Rename a character.

Works on a saved character too: the name is the one thing that can still change. Its @mention stays the same, so directions and prompts already written still find it.

Your connected YouTube channels.

The channels connected to your account. can_upload says whether a channel can take an upload, and uploads_per_channel_per_day is the daily limit per channel.

Publish a video to YouTube.

This posts to your channel, so confirm the channel, time, title and privacy first. A later time uploads it when that time comes, at the privacy you chose. Each channel can take a set number of uploads a day. Poll GET /youtube/posts for the YouTube link.

Your YouTube posts.

Newest time first. A published post has its youtube_url, and a failed one says why.

Change, cancel or retry a post.

Only a post that has not gone out: edit a scheduled, failed or cancelled one, cancel a scheduled one, retry a failed one (at once, unless you send a new publish_at), or publish a scheduled one now. A cancelled post is scheduled again with POST /youtube/posts. A published video is changed in YouTube Studio instead.

Channel and video analytics.

Views, watch time, subscribers and click-through over the last days, against the period before, plus your top videos. From YouTube’s own reporting, which runs a day or two behind.