Black Friday deposit bonus·Top up $50+ and get $10 · Top up $100+ and get $25·Ends November 30Top up →
Video

Live video API that streams an endless shot

POST /v1/videos/live streams ~5 s video chunks with sound that carry on from each other. Steer it with prompts or a director model. Billed per second.

Last updated 2026-11-05

Prepaid, no subscriptionREST or MCPOne key for models, data and agents
live video$ curl -N https://lovelyrouter.com/v1/videos/live -d '{"model": "lt… event: session · event: chunk 0.mp4> POST …/prompt "she turns to the window" event: chunk 1.mp4 · event: chunk 2.mp4 ✓ streaming · steered live
Black Friday

Black Friday deposit bonus

Every qualifying top-up earns free credit, on top of any other offer.

Ends November 30

Deposit $50
+$10
20% bonus
Deposit $100
+$25
25% bonus
Sign up and claim

A live video doesn't end when the render does. POST /v1/videos/live opens a stream of server-sent events that delivers a new mp4 chunk of about 5 seconds every few seconds. Each chunk carries on from the last second of the one before, the motion included, so played back to back they make one continuous shot with its own sound. You change what happens next while it plays: send a new prompt, or let a director, a chat model, write every next shot and take your directions. It costs $0.019 per second of video made, with a minimum of $0.42 per session.

Hand it to your AI agent

Copy this prompt into Claude Code, Cursor, Codex or any coding agent. It tells the agent where the docs are and how to use your API key.

Stream live AI video with the Lovely Router live video API and steer it while it plays.

- Docs (read first): https://lovelyrouter.com/docs/models/live-video.md
- POST https://lovelyrouter.com/v1/videos/live opens a server-sent event stream: `session`, `status`, `chunk` (each with an mp4 URL to play in order) and `end`. Steer the next chunk with POST .../<id>/prompt (or .../direct with a director), stop with .../stop; closing the stream ends it too.
- Needs `consent: true` (read the licence linked in the docs first). Billed per second streamed with a minimum per session; one open session per account.
- Auth: read my API key from the LOVELY_API_KEY environment variable and send it as `Authorization: Bearer $LOVELY_API_KEY`. Never print or commit the key. If it is not set, ask me to create one at https://lovelyrouter.com/app/keys.
- Billing is prepaid: every call is charged to my balance, so test with a small request before running anything large.
- Every docs page is plain Markdown; the full index is https://lovelyrouter.com/llms.txt.

My task: <the first frame or scene, what should happen and how you will steer it>
01

What streams back

EventWhenWhat it carries
sessionFirstThe session's id, seconds_per_chunk, price_per_second and price_min
statusWhile a GPU starts, the director writes or a chunk takes longstate (starting, directing, rendering, director_error), waiting_s
nextThe director has written a coming chunk's promptindex, prompt, by
chunkOnce per ~5 s of videoindex, url of the mp4, the prompt it followed, by (director or you), charge_usd so far
endLastreason (max_seconds, stopped, client_left, insufficient_quota, error), seconds, charge_usd

Chunk URLs need no API key, each on its own unguessable path, and are kept for 7 days.

02

Start a stream

010203

ltx-live (LTX-2.5) renders 704×384 landscape or 384×704 portrait at 24 fps, with sound, faster than real time: a 5 s chunk takes about 3 s. Every request sets consent: true to accept the LTX terms.

curl -N https://lovelyrouter.com/v1/videos/live \
  -H "Authorization: Bearer $LOVELY_API_KEY" \
  -H "content-type: application/json" \
  -d '{"model": "ltx-live", "prompt": "A fox trots through fresh snow at dusk, the camera tracking beside it", "seconds": 120, "consent": true}'

seconds (5 to 600, 60 by default) is the most the stream makes; it also ends when you stop it or close the connection. Prompts take up to 1,000 characters: say what is on screen, what happens, the light and the camera. Short prompts render fastest.

Playing it back. The first chunk can take up to about a minute when no GPU is warm, with status events meanwhile. After that, chunks arrive a little faster than they play. Buffer two, then play each as the previous one ends; two <video> elements that swap play it without gaps. To try it without code, open the playground in live mode.

03

Steer it while it plays

LIVE_ID is the id from the session event, or the x-lovely-request-id header.

Change the prompt

POST /v1/videos/live/LIVE_ID/prompt with {"prompt": "…"}. The next chunk to start follows it word for word; the one rendering now keeps its own.

Give a direction

With a director, POST /v1/videos/live/LIVE_ID/direct with {"message": "…"} works the direction into the next prompt it writes, up to 100 directions per session.

Stop

POST /v1/videos/live/LIVE_ID/stop ends the session after the chunk in progress.

Look back

GET /v1/videos/live/LIVE_ID returns the session: every chunk URL so far, the directions, the status and, once ended, the charge.

cURL
curl https://lovelyrouter.com/v1/videos/live/LIVE_ID/prompt \
  -H "Authorization: Bearer $LOVELY_API_KEY" -H "content-type: application/json" \
  -d '{"prompt": "It starts to snow harder; the fox stops and looks up"}'
04

The director

{}

Set director to a chat model's id, or to {"model": "…", "instructions": "…"} with notes on style and story (up to 2,000 characters). While one chunk renders, the director writes the next as if told "continue": from your prompt, the last 15 shots and anything you said since. It keeps the same people, place and look, moves the action on a little each shot, says what the camera does and what we hear, and writes spoken lines in quotes with who says them, about 10 words per chunk. A direction that arrives after a prompt was written gets that prompt rewritten before its chunk starts, so your directions show within seconds.

cURL
curl -N https://lovelyrouter.com/v1/videos/live \
  -H "Authorization: Bearer $LOVELY_API_KEY" -H "content-type: application/json" \
  -d '{"model": "ltx-live", "prompt": "A woman walks home through a rainy city at night", "seconds": 300,
       "director": {"model": "claude-haiku-5.5", "instructions": "Slow, moody, film noir"}, "consent": true}'

A fast, inexpensive model is plenty: the director has about 10 seconds per call, and claude-haiku-5.5 takes about 6. Its calls are billed as ordinary chat calls at that model's price and appear in your logs. If a call fails, the video keeps the last prompt.

05

First frame, fixed camera and voice

First frame

Send image (a file, an http(s) URL or a data: URL; PNG, JPEG or WebP up to 15 MB) to start from a picture, cropped to the size from the center. It stays the stream's reference: faces, clothes and the room are kept as they are in it however long the stream runs. Make one with the image API and pass the file.

Fixed camera

"camera": "fixed" makes the stream one locked-off shot, for a person talking to the camera, a room or a window. The camera never moves or cuts, and each chunk starts from a cleaned copy of the room so a long shot doesn't slowly lose its background.

Voice

By default the voice heard in the first chunk with speech is kept for every later chunk (keep_voice). Send voice, a clip of about 5 seconds of clear speech from one speaker, to choose it yourself; "keep_voice": false lets each chunk make its own.

Repeatable

With seed, chunk n uses seed + n.

06

What you pay

$0.019 per second of video made, and at least $0.42 per session, about 22 seconds of video, which covers starting a GPU and keeping it ready. When the session starts, the most it can cost, seconds × the price, is held from your balance; if your balance is lower, less is held and the stream ends with insufficient_quota when that is used. The hold is released at the end and the seconds made are billed as one call. A session that ends on our error pays only for the video it made, without the minimum. The director's calls are billed separately as chat. See pricing for everything else on the same balance.

ModelInput → outputPrice
LTX-2.5 Live
ltx-live
text+image→live video$0.019 per second

Live prices in USD, the same the API charges.

07

Limits

  • One open session per account; a second gets 429 until the first ends.
  • A small number of GPUs serve the model. When all are streaming, a new session gets 503 with retry-after.
  • A session runs up to 600 seconds. To keep going, start another and pass the last chunk's final frame as image.
08

Frequently asked questions

?
What is a live video API?+

An endpoint that keeps making video while you watch, instead of returning one finished clip. Here it streams ~5 s mp4 chunks as server-sent events, each continuing from the one before, and you can change the prompt at any moment.

How much does live video cost?+

$0.019 per second of video made, with a minimum of $0.42 per session. A director model's calls are billed as chat at that model's price.

How long can a live video run?+

Up to 600 seconds per session. Start a new session from the last frame as image to continue the same shot.

Does the video have sound?+

Yes. Each chunk has its own sound, made by the model from the prompt, including spoken lines. The speaker's voice is kept across chunks by default, or you can upload one.

Can I keep the same person on screen?+

Yes. A first frame stays the stream's reference, so the people and the place in it are kept for the whole stream.

What licence applies?+

The LTX terms. Every request sets consent: true, accepting them and the responsibility for following them.

For finished clips with tagged characters, see the video shots API; for a song turned into a video, music videos.

One prepaid key for models, data and agents

No subscription. Top up, use what you need, see every charge in your logs.