KinoPipe
Start for free
KINOPIPE FOR AGENCIES

The ad is approved. Now make it in four shapes, forty times a week.

Assemble an ad from the client images, a voiceover and the job text, then decline it to 9:16, 1:1, 4:5 and 16:9 from the same source. The design stays yours. The rendering is the part that should not hold an ad for a day.

Read the slideshow workflow Try a video tool
How one approved brief becomes every format
01

Send the voiceover to transcribe-video. Its word timestamps tell you how long each paragraph runs, so the images can be timed to the words rather than to a guess.

02

Turn each image into a clip of the length its paragraph occupies. The job text and the client logo go on in the same call through create_video_job, which takes an add_text operation with a position, size, colour and a boxed or outlined style, plus a watermark composition.

03

Join the clips with merge-videos, lay the voiceover over them, and burn the captions from the SRT that transcribe-video already returned.

04

Decline the finished ad with pad_to_ratio to 9:16, 1:1, 4:5 or 16:9, on a blurred or black background. Same source, one job each, no second edit.

Formats from one source
9:16, 1:1, 4:5, 16:9
Median render
8.2s for a 1080p H.264 edit
Composed edit
One FFmpeg pass, so no generation loss

Ask for the output you need.

KinoPipe gives Agencies focused tools with bounded fields. Include the source and target format in your request.

“Build the ad from these six images and this voiceover, then give me 9:16 and 1:1.”

“Put the job title top left and the client logo bottom right, in the same pass.”

“Render last week's ad at 4:5 without touching anything else.”

An idempotency key on every call means a rerun of the same brief returns the original job instead of billing a second render. That matters more when the retry is a workflow doing it than when it is a person.

KinoPipe and Agencies

Does it generate the voiceover?

No. You bring the audio, from whichever voice you already pay for, and it gets transcribed here so the images and the captions line up with it. Nothing here synthesises speech.

Is there a template editor?

No, and that is the honest limit of it. There is no canvas, no brand kit, no motion graphics. Text position, size, colour and a boxed or outlined style are parameters on a call. If your ads need a designed layout that somebody approves by looking at it, that part stays where it is today.

What does it replace, then?

The waiting. Assembly and format declensions need no judgement, and they still hold an ad for a day because somebody has to open a tool and do them. From a workflow they run in seconds, and a composed recipe compiles to a single FFmpeg pass.

Does it fit into n8n?

Yes, over plain HTTP Request nodes, which is what n8n Cloud can run since it ships no FFmpeg of its own. There is a community node and three importable workflows on GitHub if you would rather start from something that already runs.