Making videos

Every video starts with one call: POST /v1/render. This page explains the choices you make in that call. For every field and its limits, see the Create render reference.

Brief or scene plan

Send one of these.

A brief

A brief is a plain-language description of the video, up to 4,000 characters. Reignitor plans the scenes, writes the narration and starts the render.

{
  "brief": "A 30 second advert for Northwind Coffee's weekly subscription. Show the beans, the roasting, and a happy customer opening the box.",
  "kind": "ad",
  "duration_sec": 30,
  "brand_name": "Northwind Coffee",
  "category": "coffee"
}

Fields that shape the plan:

Field	Default	What it does
kind	ad	ad for an advert, training for a training video.
duration_sec	20 for ad, 60 for training	Target total length, 5 to 120 seconds.
narrator	true	false makes a video with no voice-over.
brand_name	none	The brand name the planner should use.
category	none	The product category, for example coffee or saas.
title	written for you	A title for the video.

Planning spends credits. If planning fails you get 502 PLAN_FAILED and the planning credits are refunded. If you send plan_notes back to a person, they explain anything the planner flagged in its own plan.

A scene plan

A scene_plan is your own list of 1 to 12 scenes, rendered in order, as given. Use it when you already know what each scene should show and say.

{
  "title": "Northwind launch",
  "scene_plan": [
    {
      "type": "motion",
      "role": "hook",
      "prompt": "Fresh coffee, delivered weekly",
      "script": "Fresh coffee, delivered every week.",
      "motion": { "beat": "hook" }
    },
    {
      "type": "t2v_fast",
      "role": "body",
      "prompt": "Slow close-up of coffee beans tumbling out of a kraft paper bag onto a wooden table, soft morning light",
      "script": "Roasted in small batches, a few days before it reaches you."
    },
    {
      "type": "motion",
      "role": "cta",
      "prompt": "Order at northwind.example",
      "script": "Order today.",
      "motion": { "beat": "cta" }
    }
  ]
}

Scene types:

type	What it renders
motion	Crisp animated text, numbers, steps or a logo. Use it for anything that must be readable.
t2v_fast	Generated footage of people, places or objects, from prompt.
i2v_fast	Generated footage that starts from an image: the scene’s ref_image_url, or your product photo.
s2v	A presenter talking to camera. Needs a face.
walk_talk	A presenter walking and talking. Needs a face.
footage	Your own clip from video_url, played as-is.

Each scene takes a prompt (what the camera sees, or the on-screen copy for motion) and an optional script (the narrator’s line). The full list of scene fields is in the reference.

Tip

Generated footage is poor at readable text. Put screens, prices, steps and slogans in motion scenes, and keep t2v_fast for the real world.

Brand

An API render never picks up a brand on its own, even if your account has brand kits. You choose the brand in each request:

• brand_logo_url: your logo, as a public https URL or a data URI.
• brand_colors: optional, up to 12 hex colours, such as ["#3B2416", "#F4E9DC"].

{
  "brief": "A 20 second advert for Northwind Coffee.",
  "brand_logo_url": "https://example.com/northwind-logo.png",
  "brand_colors": ["#3B2416", "#F4E9DC"]
}

Without a logo, the video renders with no logo and no end card. To say so explicitly, send brand_enabled: false.

Brand kits

brand_kit_id names one of your saved brand kits. For now, send it together with brand_logo_url. A brand_kit_id on its own is refused with 422 BRAND_REQUIRED. A kit that belongs to another account is never used.

Product photo

product_image_url is a photo of your product (a public https URL or a data URI). It keeps the product the same in every scene, instead of letting the model invent one.

Some formats sell a product and need a photo. If you set such a route (for example product-ad, unboxing, ugc-ad or app-demo) and the plan shows the product, a render with no photo is refused with 400 PRODUCT_IMAGE_REQUIRED. To render without one on purpose, send product_enabled: false.

Presenter

s2v and walk_talk scenes show a person talking. They need a face: avatar_preset_id for the whole video, or ref_image_url on the scene. A plan with a talking scene and no face is refused with 400 PRESENTER_REQUIRED. To let Reignitor use a stock presenter, send presenter_enabled: false.

Nothing to anchor the video

When you set a route and send no brand, no product photo and no presenter, and the plan would generate all of its pictures, the render is refused with 422 NO_ANCHOR. Without an anchor the model invents the product and the people, and they will not match your brand.

Add one of the three, or send product_enabled: false or presenter_enabled: false to render anyway.

Briefs are planned before these checks

With a brief, planning happens first and is charged. If the planned render is then refused by one of the checks above, the planning credits are not refunded. Send the brand, product photo or presenter with the brief, so the check passes the first time.

Look and sound

Field	Values	Default
aspect	9:16, 16:9, 1:1, 4:5, 4:3, 3:4	9:16
quality	draft, standard, premium	draft
language	a language code, such as en, fr, de	en
art_style	cinematic, stop_motion, mascot, watercolour, whimsy_3d	cinematic
music_style	upbeat-tech, cinematic-epic, lo-fi-chill, corporate-pop	set by the format
voice_preset_id	a narrator voice ID	a default voice
upscale	true or false	set by the format and quality

Higher quality and upscale cost more credits and take longer.

Training videos

Set kind: "training" with a brief to plan a training video instead of an advert. The planner writes steps and a recap, and uses motion scenes for on-screen instructions.

{
  "brief": "How to reset a password in the Acme admin panel. Three steps, then a recap.",
  "kind": "training",
  "duration_sec": 60,
  "aspect": "16:9",
  "brand_logo_url": "https://example.com/acme-logo.png"
}

Retries and idempotency

Send an Idempotency-Key header with a unique value, such as a UUID, on every render. If your request times out and you retry with the same key, Reignitor refuses the repeat and does not render or charge a second time.

• The repeat is refused, so it does not return the original job_id. Save the job_id from the first response whenever you get one.
• With a scene_plan, the repeat returns 409 DUPLICATE_REQUEST.
• With a brief, the repeat can return 402 with the message This request was already submitted. Treat it the same as a 409.
• Use a new key for each new video. Reusing a key for a different video refuses that video.

Without the header, every call is a new render and a new charge.

Watermarks

The render response has a watermarked field. Videos made on the Agency plan are not watermarked.

Output

When the job succeeds, video_url on GET /v1/job/{id} is a direct link to the MP4. Download it and keep your own copy.