Title card then clip: an Ideogram still as the first Timeline slot

Timeline 1.0 accepts a still image as a video[] source and holds it static, so a headline from ideogram/ideogram-v4.5 can be the first slot and a Wan 3.0 clip the second. Render is $0.10 per started minute, so a short title card plus one clip is one billed minute.

Step 1: make the title card

Generate the still at the output shape. Timeline output defaults to 1080x1920, so ask Ideogram for 9:16 and keep the headline in the middle band, away from the edges.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ideogram/ideogram-v4.5","prompt":"Title card, the words \"THREE TIPS\" in large cream type centered on a dark teal background","aspect_ratio":"9:16","quality":"medium"}'

Step 2: make the clip

Submit the 9:16 clip to POST /v1/videos with wan-3.0 and poll until it is completed. Take the media.sume.com URL from the job result artifacts, since Timeline reads Sume-hosted media.

Step 3: join them

Put the still first in video[], each slot with its own video[].duration (at least 0.2 s). The docs say a still source is a static hold; the job accepts a motion field on it, ignores it and reports motion_ignored, so leave motion off to keep the response clean.

audio.duration_seconds is required and sets the output length. Coverage from the video slots can stop at most 0.5 s before the end of that spine, so make the slot durations add up. audio.mode: "silence" is allowed if you only want the picture; the Timeline reference in the docs has the full request.

What it costs

Plan requests are unbilled, so check the layout first. A render reserves ceil(audio.duration_seconds / 60) minutes at $0.10 each, and the image and clip are billed when generated.

When not to do this

If the platform already shows a cover frame, a title card spends the first seconds of a Short on text. Use it when the opening line is the hook, not as decoration.

Request sketch

A plan call is a safe place to test this because it is unbilled and reports problems such as motion_ignored before you pay for a render. Use short slot durations while you check the layout, and set audio.duration_seconds to the sum of the slots.

The first slot is the title card at about two seconds, which gives viewers time to read a three-word headline. The second slot is the Wan clip at its generated length. If the clip is five seconds, the spine is seven seconds.

Keeping the look consistent

Match colors between the still and the clip. If the card is dark teal and the clip opens on a bright kitchen, the cut will feel abrupt. Either ask Wan for a similar palette in the prompt or choose a card background taken from the clip's first frame. The hold has no transition on it unless you add one in the slot, so pick a cut that looks right.