Stream starting soon screen: make an animated loop

A stream starting soon screen is a short looping video, or a still image, that you show full-screen before you go live, with words such as “Starting soon” and your channel name. Make it the size of your stream's canvas (1920×1080 for a 1080p stream, 1280×720 for 720p), and hide the loop point by ending the clip on the same frame it starts with.

With Sume that is three steps: an image-to-video clip that starts and ends on your background image, one caption job that burns the words, and a Timeline render that repeats the clip for as long as you need. Facts come from the Video generation, Video captions, and Timeline 1.0 docs and the Sume API reference, read on 2026-09-28; anything called current code is read from Sume's source.

What size should a starting soon screen be?

The same size as your stream's output, so it fills the frame without being scaled: 1920×1080 for a 16:9 stream at 1080p, 1280×720 at 720p. Timeline 1.0 renders any even size from 256 to 2,160 pixels per edge, so both work, but a 2560×1440 canvas is wider than that limit.

How do I make an animated starting soon screen with AI?

Start from a 16:9 background image at a public HTTPS URL, your own art or one you generated. Leave the words out of the picture: they go on in the next step, as text you write.

• Send the image to POST /v1/videos as both the first_frame and the last_frame, and describe calm motion that can come back to where it started. In the current catalog, every listed model takes a last frame except grok-imagine-video-1.5; Seamless loop AI video shows how to check the loop.
• Keep the clip's sound on; generate_audio defaults to what the model can do. In current code the caption job refuses a clip with no audio stream, and the Timeline step drops the clip's own audio anyway.
• When the job completes, read the clip's media.sume.com URL from GET /v1/jobs/{id}/result.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: starting-soon-loop-v1" \
  -d '{
    "model": "wan-3.0",
    "prompt": "Slow drifting clouds and soft neon light over a city skyline; the camera stays still",
    "frame_images": [
      { "type": "image_url", "image_url": { "url": "https://example.com/starting-soon-bg.png" }, "frame_type": "first_frame" },
      { "type": "image_url", "image_url": { "url": "https://example.com/starting-soon-bg.png" }, "frame_type": "last_frame" }
    ],
    "aspect_ratio": "16:9",
    "resolution": "1080p",
    "duration": 10
  }'

How do I add the “Starting soon” text?

Burn it with POST /v1/video-captions: one cue whose text is your line, with start at 0 and end at the clip's length. Cues skip speech-to-text and burn your words at those times; How to add text over a video covers where the line sits.

• In current code the default slam style prints the words in capitals and lays a light black tint over the whole frame.
• Put “Starting soon” and your channel name in the same cue: in current code slam draws every cue at the same height and ignores line breaks.
• Set design.motion.enter_seconds and exit_seconds to 0 (each takes 0–2) so the words have no entrance or exit animation at the loop point.

How do I make it run as long as my pre-show?

Loop the captioned clip in one Timeline 1.0 slot as long as the screen should run, with render.pad_mode set to loop; How to increase video length by looping explains the mechanics. For a pre-show there is no voice, so set audio.mode: "silence" with the length in audio.duration_seconds, up to 1,800 seconds (30 minutes), and set output to 1920×1080, since the default frame is 1080×1920.

• Timeline reads only your workspace's media.sume.com files. Completed Sume jobs can include public artifacts there, including the captioned clip.
• For music, add a Sume-hosted soundtrack with loop: true and gain_db at 0; the default −16 dB is a bed level under a voice. AI lofi video generator pairs the same kind of loop with a track.

curl -X POST https://api.sume.com/v1/timeline-1.0/render \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: starting-soon-30min-v1" \
  -d '{
    "audio": { "mode": "silence", "duration_seconds": 1800 },
    "video": [
      { "source_url": "https://media.sume.com/artifacts/artf_demo/starting-soon.mp4", "start": 0, "duration": 1800 }
    ],
    "output": { "width": 1920, "height": 1080 },
    "render": { "pad_mode": "loop" }
  }'

Can the screen show a live countdown?

No. A rendered file can't know when you'll actually go live, so any countdown in it is fixed. For a short one, caption a separate clip with one cue per second and the same zero enter and exit times: a caption job takes up to 200 cues, each timed within 60 seconds, and in current code its source must be 60 seconds or less with an audio stream, such as a 30-second generated clip with its sound on. Play it after the loop, not inside it, or the count restarts on every pass.

What are the limits, and what does it cost?

The clip is reserved at its model's provider list price × 1.25. Timeline is listed at $0.10 per output minute and reserves whole minutes, so a 30-minute screen reserves $3.00. Both are plus a 5.5% agent fee by default. A caption job is a fixed amount per clip of up to 60 seconds, listed on the Video captions page.