A slideshow is not a video model

A slideshow is a stack of stills with a duration per frame, optional overlay, optional music, then a render to MP4. It is not a video model. It does not invent camera moves, visemes, or a scene from a cinematic paragraph. AI slideshow maker sequences frames. Text-to-video generates a clip. Sending “a six-slide carousel about morning habits, cinematic, 4K, native audio” to a video row is a new film that happens to be about a list. You will not get six cards you can reorder.

Fifteen of nineteen public boards on /slideshows are six frames at 36 credits. That catalog is a scene graph — slide count, cost, prompt, aspect — not a vibe. Recreate a graph. Do not ask Kling to be a carousel.

Frames and timing are the object

The object has:

• N stills, in an order you can edit.
• A hold (duration_per_image on convert is one number for the set, not per-slide storytelling).
• Optional overlay captions (copy stamped on top) or captions baked into the image by an image model.
• Optional voiceover_url / music_url that already exist. Convert does not write a track.
• An aspect. Public listings are 9:16 reels. Set the ratio on the stills, not later in a video prompt.

What it does not have:

• A duration ladder like 4s / 6s / 8s on a generate row. Those seconds are a clip. A slideshow’s seconds are N × hold.
• Native audio from a video model. A bed is a file you attach.
• Motion control, lipsync, first-last interpolation. Those are other slugs.

Turn photos into a slideshow arranges your images. Generate a slideshow from a prompt plans stills, generates or curates them, stamps overlay if you asked. Priced per image generated, shown before you confirm. That meter is image credits, not per-second video. A six-pack at ~6 credits a frame is a different bill from a 40-credit Veo shot.

Pace is one number on convert. If you needed a six-second push-in on one product still, that is image-to-video, not a 3-second crossfade across eight cards. Pick the job that matches the motion you actually want.

Where the stills come from

Three honest sources:

Your photos. Order is the order you give. Overlay captions are optional copy the agent writes and stamps. The images are already paid (you shot them). Do not run a video model “to make them feel like a slideshow.” Convert will do that with a hold and a transition.

Generated stills. Each slide is an image generate: Ideogram if the slide is a title, Seedream if it is a dense board, Recraft if it is UI, a cinematic stills row if it is a plate. Proof type on the still. A video model will not letter slide 4 for you.

A public listing. Recreate the locked stack — slide count and the estimate on the row. Age-of-25 is 11 frames at 242 credits. Improve Yourself is 6 at 132. Those are not 36-credit six-packs. Asking a planner to “do Age of 25 cheaper in six slides” is a different board that happens to share a title.

Wrong source: a single cinematic generate you will “cut into slides.” You will cut a film. You will not have six stills you can swap. Change a slide is a delta on a board. It is not a trim on a Veo take.

Lettering in the frame is a stills specialist, not a slideshow feature. If slide 1 is a title card, generate the title on Ideogram or Seedream, then put that file on the board. Overlay is for lines you will restyle. Baked-in image type is for letters that are the picture, and that style is async (slides land over 30–90 seconds). Ask for the one you mean.

Convert is a render, not a generate

Turn a slideshow into a video is convert_slideshow_to_video. It needs a slideshow_id. It does not write slides, does not generate a cinematic clip from a sentence, and does not compose random stills you never put on a board. Using it as a text-to-video shortcut spends a render on an MP4 you will throw away — and the slide images were already paid if they were generated.

Fix the board first. Reorder, replace slide 3, stamp overlay. Then convert. A second convert after you fix slide 3 is a second process. Rendering is a flat processing step, not a per-model generation. The exact cost is shown before it runs. Cheap compared with a video model, still not free.

Audio URLs must already exist. Generate VO on a TTS row. Generate a bed on a music row. Then attach. Convert will not invent a chorus.

use_edited_images is whether stamped overlays come along. Forget that flag and you render bare originals.

Stitching loose images into one video without a slideshow project is also not convert. That is a compose/stitch job. Convert is the carousel you already named.

When you actually wanted a video model

Use a video row when the motion is the brief: a product turn, a talking mouth, a first-last interpolate, a 30-second single take. Those seconds are generated. They are not a hold on a PNG.

Use a slideshow when the brief is a list, a kit, a recap, a carousel of claims, a wardrobe taxonomy, six labelled fails. The viewer is reading cards. A cinematic generate of “six tips” will morph the tips. You cannot fix tip 4 without rerolling the film.

Mixed pipelines are allowed and should be named:

• Slideshow of stills, then convert, for the carousel cut.
• One hero still sent to image-to-video for a motion insert, edited onto the timeline — not “the slideshow model.”
• Captions on the converted MP4 if someone speaks. Auto captions transcribe speech. They do not read overlay. Overlay is already in the stills.

Article 50 still applies if slides contain realistic synthetic people and you publish them as authentic-looking media in the EU. A six-frame kit of products usually does not. A generated testimonial face does. Article 50 disclosure is the copy when the rule applies.

When the brief is frames plus timing, build a board. Sign up and run the slideshow maker as stills in a sequence, not as a video model with extra steps.

FAQ

Can I prompt a video model to output a six-slide carousel?

You can prompt it. You will get a clip, not six editable stills. Build N frames on image rows or uploads, put them on a board, convert. That is the carousel.

Is convert_slideshow_to_video a Veo generate?

No. It renders slideshow_id with a hold, a transition, and optional existing audio. Scene generation is a different job. Movie mode is a different job.

Why are some six-slide boards 36 credits and some 132?

The public catalog is not one meter. Fifteen of nineteen listings are 6 images / 36 credits. Improve Yourself is 6 / 132. Read the row. Recreate that graph. Do not average the shelf.