Wan 2.5 I2V is silent; T2V on the same rate carries audio

Wan Video 2.5 I2V and T2V both bill 15 credits at 1080p for 5s or 10s; the image-to-video row has audio false, the text-to-video row has native audio — same family, different meter flag.

Teams treat “Wan 2.5” as one button. It is two slugs. The sticker is the same: 15 credits, 1080p, 5s or 10s. The soundtrack is not. Image-to-video comes back mute. Text-to-video comes back with a mix the model wrote. Pick the row for the input you actually have, then write or refuse the sound on purpose.

Same 15 credits, two meter flags

Wan Video 2.5 Image to Video is high-quality image-to-video with Wan 2.5. Category is image-to-video. It requires an image. audio is false. Durations: 5s and 10s. Ceiling 1080p. Display price 15 credits.

Wan Video 2.5 Text to Video is the sibling: high-quality text-to-video with Wan 2.5. No still required. audio is true. Same durations. Same ceiling. Same 15 credits.

That is the whole fork. Not a quality tier. Not a discount flag. Same family, different meter: one row is silent motion on a plate you already approved, the other is a 5s or 10s scene with a soundtrack from a paragraph. Pasting a T2V prompt into the I2V slug does not unmute it. Uploading a still to the T2V slug does not make the row honour that plate as a contract.

Fifteen credits is cheap next to silent 1080p rows that bill 60+. That is the trap on both slugs. Teams treat the I2V pass as a disposable picture test and then wonder why there is no bed. Teams treat the T2V pass as a disposable picture test and then ship an accidental soundtrack. The flag is the generate. Read it before you confirm. The I2V versus T2V fork in general is which to use. For Wan 2.5 the fork is also two URLs.

I2V is silent and needs a still

The still is the brief. Composition, product, type, the face — all of that is decided before this row. Wan 2.5 I2V’s job is motion on that file for 5s or 10s at up to 1080p. It is not a second chance at the layout.

AI image-to-video is the door. Image in, clip out. No still, no generate. A paragraph where the image should be is a wrong label. Every second of the 5s or 10s is waste. audio is false. Prompting a voice line here does not buy dialogue. It may smear the plate with painted words. If you need a line, generate speech on a TTS row and lay it in the editor after this pass. If you need a whoosh, that is a sound-effect generate, not a clause on this slug.

1080p is the ceiling. Quality chips include 480p, 720p, and 1080p. Pick the quality the still can support. A soft plate at 1080p is a sharp view of softness. Motion levels Medium and High are extras on a locked still, not a substitute for the still. High motion on an unapproved composition is 10s of confident drift. Medium on an approved product still is often the whole job. Styles listed: Cinematic, Documentary, Abstract. Those shift motion register. They do not unmute audio. They do not remove the image requirement.

Correct I2V label: approved still, 5s or 10s, up to 1080p, silent, 15 credits. Wrong label: no image; expecting a 15s or 30s chip; expecting native audio; expecting the model to invent the pack.

T2V carries audio from text

The T2V row does not require an image. It invents the frame. If a specific pack or face must appear, lock it on a stills row first, then change to I2V. Do not describe the SKU in adjectives and hope Wan 2.5 T2V invents the label you would print.

Native audio means the 1080p file arrives with a track. Write the room and the line the way you write the camera. Quote dialogue. Name the hiss, the crowd, the none. A picture-only prompt still returns a soundtrack the model chose. The AI video generator will not mute this row because you forgot.

Aspect options are 16:9, 9:16, and 1:1. Quality settings include 480p, 720p, and 1080p. Frame rates listed are 24fps, 30fps, and 60fps. Motion levels listed are Low, Medium, and High. None of those knobs turn audio off. Documentary still gets a mix. Use 5s when the hook is a single action. Use 10s when the action has a land. There is no 4s, no 6s, no 8s, no 12-second hold on either Wan 2.5 video slug. If the job is a silent product turn, T2V is the wrong row even though the credit sticker matches. Silent coverage is the I2V slug, or a later mute in the editor on a file that already contains a bed you may have to fight.

Pick the row for the input you have

If you have an approved still and you need motion without a mix, you are on Wan 2.5 I2V. Fifteen credits. Mute file. Do not write a VO into that prompt.

If you have a paragraph and you need a 5s or 10s scene with diegetic sound, you are on Wan 2.5 T2V. Fifteen credits. Write the cue. Do not “fix it in captions.” Captions will transcribe the line you did not write.

If you have an approved still and you need a mix, do not assume I2V grew a soundtrack because the sibling has one. Score after picture lock, or pick a native-audio image-to-video row that actually lists audio true. Same family is not the same flag. If you have neither a still nor a sound brief, stop. Fifteen credits of high-quality wrong is still wrong. Lock the plate, or write the cue, then pick the slug that matches. Do not stitch the two slugs. I2V then T2V is two bills and two identities. T2V then “match this still” is still a generate of a cousin. One input. One row. One flag.

FAQ

Does Wan 2.5 Image to Video need a still?

Yes. Category is image-to-video. If you only have text, use Wan Video 2.5 T2V, or lock a still first.

Is there a soundtrack on the I2V row?

No. audio is false. 5s or 10s, up to 1080p, 15 credits, silent. Mix speech after. The T2V sibling at the same rate has native audio.

If I do not mention sound on T2V, is it silent?

No. Native audio is on. A picture-only prompt still returns a soundtrack the model chose. Write the cue, or write silence, and assume you may still need a replace pass.

Can I generate 15s or 30s on either Wan 2.5 video slug?

No. Durations are 5s and 10s on both. A longer need is a different row or an extend pass, not a prompt trick on these slugs.