Discord
I2V
Wan 2.6 image-to-video via wan-video/wan-2.6-i2v
1
Camera verb per Wan 2.6 clip
3
Presets: cinematic, portrait, wide
14
Prompt chips below for quick tests
Each row shows the Wan 2.6 input keyframe beside a motion pass. Upload a locked still, pick one camera preset, and describe motion only — same subject, same light, only time moves.
Cinematic dolly
Slow cinematic dolly in over misty cliff path at golden hour, gentle cloud drift, static horizon, one camera verb only.
Input still
Wan 2.6 I2V
Portrait push-in
Subtle head turn and soft breathing, city window bokeh unchanged, gentle push-in, face stays sharp.
Input still
Wan 2.6 I2V
Wide neon alley
Slow forward drift through neon rainy alley, puddle reflections shimmer, static camera height, ambient rain bed.
Input still
Wan 2.6 I2V
Wan 2.6 is the still-to-motion checkpoint in the Wan video family — this generator defaults to wan-video/wan-2.6-i2v.
| Checkpoint | Best for | Honest note |
|---|---|---|
| wan-video/wan-2.6-i2v | Wan 2.6 still-to-motion, hero keyframes | Default on this page — upload a still, preset handles camera grammar |
| prunaai/wan-2.5-i2v-fast | Fast Wan I2V iteration after brief is locked | Swap in dropdown when you need speed over 2.6 identity |
| prunaai/wan-2.5-t2v-fast | Text-only briefs without a reference still | Use Text to Video when you have no keyframe upload |
Tap chips, copy the line, paste after your preset.
Wan 2.6 image-to-video
Volume first — promote only the winning clip to slow polish.
01
Agree on one camera move and one subject verb before any Wan 2.6 credits spend.
02
Generate three Wan 2.6 passes with one changed clause — star the cleanest horizon.
03
Check foot sliding on phone OLED before paid social — downscaled previews hide drift.
Wan’s official page and ranking generators lead with broad model capability. Voor focuses this route on the actual I2V controls available here and a shot-planning method you can repeat.
Use a clean still with a readable subject, plausible depth, and room for the camera direction you want.
Choose dolly, orbit, crane, handheld, or locked-off language instead of stacking conflicting moves.
Separate what the subject does from what the camera does so the model can stage both clearly.
On this I2V endpoint audio is optional background sound; it does not become the motion driver.
Evidence notes
Practical camera language, motion boundaries, prompt iteration, audio handling, source-frame review, delivery checks, and model-fit decisions make the shot reproducible.
The demos above are still→motion pairs: a locked keyframe plus one camera verb. On Voor AI this page defaults to wan-video/wan-2.6-i2v — upload a still, pick a preset, describe motion only.
Strong briefs name subject, one camera move, and one lighting note. If the clip drifts, change camera or subject — not both in the same reroll.
How we made the examples. Cinematic / portrait / wide plates were generated as stills, then animated with a single motion instruction so you can judge identity hold. Review hands, on-screen text, and foot contact on a phone OLED before paid social — downscaled previews hide sliding.
Compare Wan Animate for related I2V workflows and Image to Video AI for the full model dropdown. Use this page for volume previz; promote only the winning clip to slower polish.
Wan 2.6 is Alibaba's image-to-video model — upload a keyframe, describe camera and motion, and get a short clip. On Voor AI you get presets, example prompts, and a generator wired to wan-video/wan-2.6-i2v.
Pick a preset, add your subject, generate. Presets prepend camera language so you describe only the variable part.
New accounts receive 3 welcome credits. The generator shows the credit estimate before you run.
This page defaults to wan-video/wan-2.6-i2v in the wan-video collection — upload a still, pick a preset, describe motion.
Yes under your Voor plan terms — verify likeness and trademark rights separately.
Presets encode Wan camera grammar, copyable examples show what works, and the generator defaults to wan-video/wan-2.6-i2v instead of a blank prompt box.
One camera move, one lighting note, explicit preserve lists for edits.
You get generations from the model — pages teach the brief; you review and iterate.
FAQ below, scene ideas, and linked Voor tools at the page bottom.
Generate three Wan 2.6 variants per brief, pick the cleanest horizon, promote only that clip to slow polish — Wan 2.6 I2V is for volume, not final delivery on every reroll.
Name 9:16 or 16:9 in the Wan 2.6 prompt when placement is fixed — presets bias horizon lock; your crop note prevents awkward reframes for Shorts versus wides.
Batch three Wan 2.6 variants per brief, pick the cleanest horizon, promote only that clip to slow polish — Wan 2.6 I2V is for volume before final voiceover and color grade.
Shared doc of approved Wan 2.6 clauses — subject and light notes reuse across SKU videos while camera preset swaps per placement. Version filenames with date for client timelines.
Hands, on-screen text, and thin logos need review at full size after every Wan 2.6 export — reroll one clause instead of accepting a close-enough Wan 2.6 pass for paid social placement.
Storyboard one camera arrow and one subject verb before Wan 2.6 generation — stakeholders agree on motion before credits spend on volume tests that never ship to the client timeline.
Version Wan 2.6 exports with date and prompt hash — editors trace which pass landed when the producer asks for a rerun on the same SKU hero without reopening the whole brief deck.
Upload a still, pick a Wan 2.6 camera preset, generate image-to-video with wan-video/wan-2.6-i2v on Voor AI.
Open Wan 2.6 generatorWan 2.6 image-to-video on Voor AI — upload a keyframe, pick a Wan 2.6 camera preset, generate with wan-video/wan-2.6-i2v.
