Video to video in Telegram

Two models in the bot accept a reference video. Here is how video to video works in Telegram, and the billing rule that catches people out.

Two of the five models, Wan 3.0 and Seedance 2.5, accept an existing clip as a reference and generate from it. That covers restyling something you already shot, or carrying a motion you like into a new scene. It is also the one place in the bot where the pricing arithmetic changes, so it is worth reading before you attach anything.

The billing rule, stated plainly

The reference clip’s own seconds are added to the output’s seconds, and the total is what you pay for.

Attach a 10 second video, ask for a 10 second result, and you are billed for 20 seconds. On the workhorse at 720p that is 10 times 20, so 200 credits rather than the 100 the length picker might suggest at a glance. This is the provider’s rule and we pass it through unchanged rather than quietly marking it up elsewhere. Every rate it multiplies is on the pricing page.

Photos behave completely differently. Attach as many stills as the model allows and the price does not move at all, which is what image to video covers.

The 30 second ceiling

Input plus output cannot exceed 30 seconds on either model. The useful consequence is that the length picker narrows itself once a video is on the draft: attach a 20 second reference and the options above 10 seconds disappear, because they would breach the cap. You cannot accidentally order something the model would refuse.

When it is worth it

Restyling is the honest use case. If you have footage and want it in a different look, this is the shortest route to it, and no editor is involved.

Extending a clip is not. Because the input is billed too, feeding a 10 second clip back in to get 20 seconds costs the same as 30 seconds of fresh generation while giving you less control. Generate the length you want in one go.

Getting started

Send the clip to the bot with your instruction as the caption, the same way you would attach a photo, then pick the model. Only the two that accept video will offer the option, and the others will tell you they cannot take it rather than silently ignoring the attachment.

Frequently asked questions

Which models accept a reference video?

Two of them, the everyday workhorse and the quality tier. The other three take text and photos only.

How is an attached video billed?

Its own length is added to the output's. A 10 second reference feeding a 10 second result is charged as 20 seconds at the rate for the model and resolution you picked.

Is there a limit?

Yes. Input and output together cannot exceed 30 seconds on either model, and the length picker narrows on its own once a video is attached, so the ceiling is enforced rather than left to you.

How many videos can I attach?

Up to five on the workhorse and up to ten on the quality tier. Every attached second counts toward the bill and toward the 30 second ceiling.

Can I use video to video to make a clip longer?

Not usefully. Because the input is billed as well as the output, feeding a clip back in to extend it costs more than generating the longer version outright.

zosee is an independent product. Not affiliated with, endorsed by, or sponsored by Telegram, xAI, Alibaba, Kuaishou, Google, OpenAI or Runway. Model names are trademarks of their respective owners.

Make a clip in Telegram

Join the ideas channel and claim a one-off bonus sized to exactly one short clip on the entry model. Credits never expire.