Every frame,uncut.
The full Wan 3.0 video model, without the content filter that cuts your scene in half. Text, images, video and audio in. Thirty seconds of cinema with sound out — up to 4K.
- T2V · I2V · R2V
- 2–30 s
- Native audio
- 480P → 4K
- prompt_extend
01/Director's note
Standard filters say no to a lot of cinema. Horror that actually frightens. A fight that lands. Intimacy shot with taste, not euphemism. Wan 3.0 Spicy runs the same Wan 3.0 model — and leaves the final cut to you.
02/Dailies
Thirty seconds is a scene, not a clip.
Native dialogue with lip sync, multi-shot storyboards, first-and-last-frame control and up to twenty references in one request. These takes are official Wan 3.0 outputs — Spicy routes the same model without the filter.
Take 01 · R2V · 15s · Image + video + audio
A man by a rain-streaked window, warm lamp inside, neon outside, on the phone: “Really? That sounds great… I miss you so much.”
Take 02 · I2V · 20s · 9:16 · Timestamped
(0:00–0:03) Medium shot, a swordsman in a bamboo hat and ink-black robe — two ink-silhouette assassins close in through the bamboo.
Take 03 · Edit · 5s · Dialogue swap
Change his line to: “The deal is done. Now… we disappear.”
Take 04 · R2V · 24s · Three references
Reality and simulation collide in a dojo standoff, delivered in Japanese with matching lip sync.
Take 05 · T2V · 30s · One take
A thirty-second single take on a film set: a director whip-pans between lines while the lead actor's expression swings from tears to laughter.
Take 06 · Extend · 20s · Backward
A 19th-century palace ballroom in warm gold, extended fifteen seconds backward into the waltz that led here.
Samples: official Wan 3.0 outputs published by Alibaba Cloud Model Studio. Hover to look closer.
03/Three ways in
One endpoint. Whatever you bring to set.
Send a prompt, a pair of keyframes, or a box of references. Use a scenario model, or the all-in-one model that reads your materials and picks the mode itself.
wan3.0-spicy-t2v
Text to video
Write the scene — camera, light, dialogue, timing — and get up to thirty seconds with a matching soundtrack.
prompt · duration · resolution · aspect_ratio
wan3.0-spicy-i2v
Keyframes to video
Pin the first frame, optionally the last, and let the model direct everything in between.
image_urls — first frame · optional last frame
wan3.0-spicy-r2v
References to video
Up to 10 images, 5 videos and 5 audio clips — or a PDF or web page — cast by name in your prompt.
reference_images · reference_videos · reference_audio · file · link
Casting by reference — hover a name
picks up and starts a slow country song. walks past carrying , lays them on the table and says: “The sun is perfect today.”
Image 1
Image 2
Image 3- Video 1
04/Two cuts
Pick the print. Pay by the second.
Spicy renders 480P to 1080P. Pro Spicy renders 1080P, 2K and 4K for the big screen. Both bill one way: rate × (reference video seconds + output seconds). Audio on or off, the price is the same.
Estimated cost
$3.00
(0s reference + 10s output) × $0.300/s
- 1080P$0.200 / s
- 2K$0.240 / s
- 4K$0.300 / s
Live list prices from the APIPod catalog. Failed tasks are refunded. Smart duration (-1) is pre-charged at 30s and settled to the actual length once the video is ready.
05/On the call sheet
One request from idea to dailies.
The same /v1/videos/generations endpoint as every APIPod video model. Async by design: submit, poll, download.
- prompt_extend: false keeps the model on your literal words — no polite rewrites.
- Images accept URLs or Base64; video and audio references take URLs.
- duration -1 lets the model choose the length — pre-charged at 30s, then settled to the actual cut; 2–30 for exact lengths.
- One API key across Wan, Seedance, Veo, Kling and every other APIPod model.
curl https://api.apipod.ai/v1/videos/generations \
-H "Authorization: Bearer $APIPOD_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "wan3.0-pro-spicy-t2v",
"prompt": "Slow dolly down a flooded Victorian hallway at 3 a.m., one flickering gas lamp, a figure in a soaked nightgown turns her head too slowly. Cold blue-green grade, heavy film grain.",
"resolution": "4K",
"aspect_ratio": "16:9",
"duration": 10,
"prompt_extend": false
}'
# → { "code": 200, "data": { "task_id": "…", "status": "pending" } }
curl https://api.apipod.ai/v1/videos/status/$TASK_ID \
-H "Authorization: Bearer $APIPOD_API_KEY"06/Script library
Genres the filter used to refuse.
Starting points for mature, cinematic work. Copy one, change the cast, roll.
01·Horror15s · 16:9
Slow dolly down a flooded Victorian hallway at 3 a.m., one flickering gas lamp, water rippling as something unseen moves beneath it; a girl in a soaked nightgown turns her head too far, too slowly. Cold blue-green grade, heavy film grain, low drone and dripping water.
02·Action10s · 16:9
Handheld 35mm, a bare-knuckle fight in a rain-soaked alley under sodium streetlights; sweat and blood spray in slow motion on the knockout punch, then real-time chaos as sirens close in. Desaturated orange, shallow depth of field.
03·Neo-noir8s · R2V
Image 1 as a detective in a trench coat, cigarette smoke curling through venetian-blind shadows; slow push-in as he says, “Everybody lies. The dead just do it quieter.” Black-and-white with a single red neon accent, jazz saxophone.
04·Dark fantasy20s · One take
Low-angle crane up a cathedral of bones; a hooded witch-queen raises a blackened crown as ravens explode from the rafters. Candle-fire rim light, crimson and charcoal palette, a choir swelling.
05·Fashion film10s · 9:16
Golden-hour light through sheer linen curtains, a model in a black silk slip dress turns slowly at a vintage vanity; close-ups of earrings, lipstick and a confident over-the-shoulder glance. 85mm, soft bloom, slow-motion fabric, ambient piano.
06·War drama30s · One take
A trench line at dawn, 1917; a long tracking shot follows a young soldier sprinting through smoke as shells erupt and mud rains down, never cutting, ending on his face as the whistle blows. Muted sepia, visceral sound design.
07·Satire12s · 16:9
Mockumentary press briefing: a pompous minister in an oversized sash insists “the economy has never been stronger” while the podium slowly sinks into the floor. Static wide, deadpan cuts to reporters' faces, bright broadcast lighting.
08·Crime thriller15s · Multi-shot
Shot 1 (0–5s): overhead of a briefcase of cash on a motel bed. Shot 2 (5–10s): close-up, a woman loads a revolver, hands steady. Shot 3 (10–15s): she says, “You were never supposed to find me,” and the door bursts open. Neon pink and teal, rain on glass.
07/House rules
Fewer filters. Not zero rules.
Relaxed moderation is for adult creative work. A few lines are never crossed — by the model or by us.
- Adults only
- Spicy models are for users 18 and over. By calling them you confirm you meet the age of majority where you live.
- No minors, ever
- Any sexualized depiction of a minor — real, fictional, or implied — is strictly prohibited.
- Consent for real people
- Don't generate intimate or defamatory content of a real, identifiable person without their explicit consent.
- Follow your local law
- You are responsible for how outputs are used and distributed under the laws that apply to you and your audience.
- Regional availability
- Spicy models and their pages are unavailable in mainland China, Hong Kong and Macau; API requests from these regions are declined.
- Your cut, your rights
- Content you generate belongs to you. APIPod and the upstream provider claim no ownership over outputs.
08/Questions from the floor
Before you roll.
What is Wan 3.0 Spicy?
Wan 3.0 Spicy is Alibaba's Wan 3.0 video model served on a relaxed-moderation route for adult creative work. It keeps the Wan 3.0 contract — text, keyframe and multi-reference to video, 2–30 second clips with native audio — and is available as wan3.0-spicy-t2v, -i2v, -r2v, or the all-in-one wan3.0-video-spicy.
How is it different from standard Wan 3.0?
Same model family and request format. The difference is the upstream moderation stance: Spicy accepts mature themes the standard route rejects. It also supports prompt_extend, and is priced slightly above the standard model.
What does Pro Spicy add?
Pro Spicy (wan3.0-pro-spicy-t2v / -i2v / -r2v and wan3.0-video-pro-spicy) renders at 1080P, 2K or 4K instead of 480P–1080P. Everything else — modes, references, duration, audio — is identical.
What does prompt_extend do?
By default the model rewrites short prompts to improve motion and cinematography. Set prompt_extend to false to keep generation close to your literal text — useful when every word in the prompt matters.
How is a request billed?
Rate per second × (reference video seconds + output seconds). Reference videos are measured when you submit; a request without reference video is billed on output length only. Smart duration (-1) is pre-charged at 30 seconds and settled to the actual length once the video is ready. Failed tasks are not charged, and turning audio off does not change the price.
Where can I use it?
Everywhere APIPod operates except mainland China, Hong Kong and Macau — Spicy pages and API requests are unavailable in those regions. Spicy models are for users 18 and over.
Who owns what I generate?
You do. Generated content belongs entirely to you; neither APIPod nor the upstream provider claims ownership of outputs.
Quiet on set.
Roll camera.
Create a key, pick a cut, and send your first scene. Same API as the rest of APIPod — just fewer people telling you no.