Wan 3.0 Spicy Is Live: Uncensored Video Generation Upgraded

Wan 3.0 Spicy is now available on Siray, filling the structural gap in uncensored video generation. The new model provides text-to-video, image-to-video, and reference-driven variants — addressing the i2v-only constraint that defined the previous tier. Here is what changed and what the three endpoints can do.

What Changed: From i2v-Only to Full Coverage

The previous tier, alibaba/wan-2.7-i2v-spicy, handled image-to-video only. That meant every uncensored video started from a still image — text-to-video workflows required a separate image generation step first.Wan 3.0 Spicy removes that constraint by providing three endpoints:

• alibaba/wan-3.0-t2v-spicy — text-to-video
• alibaba/wan-3.0-i2v-spicy — image-to-video
• alibaba/wan-3.0-ref2v-spicy — reference-driven (supports up to 10 images, 5 videos, 5 audio tracks as multi-modal input)

This is the first Wan Spicy tier to offer t2v alongside i2v. It also introduces ref2v, which accepts multiple reference inputs to guide generation — useful for maintaining consistent style, character, or motion across clips.

Specs: Resolution, Duration, and Parameters

All three Wan 3.0 Spicy endpoints share the same core output parameters:

• Resolution: 480p, 720p, or 1080p
• Duration: 2 to 30 seconds (integer values)
• Aspect ratio: adaptive, 16:9, 9:16, 1:1, 4:3, 3:4
• Audio: audio_enable parameter (boolean, optional)

The 30-second upper limit ties with Seedance 2.5 Spicy as the longest single-call generation available in the uncensored tier on Siray. The 2-second lower limit is shorter than Wan 2.7 Spicy, which started at 2 seconds but capped duration at 15 seconds.

One parameter difference to flag if migrating workflows from Wan 2.7 i2v Spicy: Wan 3.0 Spicy does not include negative_prompt or prompt_expansion_enable parameters. The 2.7 tier had both, with prompt_expansion_enable defaulting to true (the model would automatically rewrite and expand prompts before generation). That expansion behavior is not present in the 3.0 Spicy API schema — what you pass as prompt is what the model receives.

The ref2v Variant: Multi-Modal Reference Input

alibaba/wan-3.0-ref2v-spicy distinguishes itself by accepting arrays of reference material:

• images — up to 10 image URLs (can be data URLs or remote URLs)
• videos — up to 5 video URLs
• audios — up to 5 audio URLs

This supports workflows where generation needs to maintain consistency with existing assets — for example, generating multiple clips that share the same character design, environment style, or motion signature. The i2v variant offers a simpler interface with a single required image parameter and an optional end_image parameter for first-frame/last-frame control.

What This Means for Uncensored Workflows

The structural upgrade is straightforward: text-to-video for uncensored content no longer requires chaining through an image generation API first. Creators working from text prompts can call wan-3.0-t2v-spicy directly.

For workflows that already rely on reference images or need precise first-frame control, wan-3.0-i2v-spicy and wan-3.0-ref2v-spicy are the paths forward. The ref2v variant's multi-input capability is particularly useful for batch generation where visual consistency across clips matters — something that single-image i2v cannot guarantee without external style-transfer steps.

One operational note: while Wan 2.7 i2v Spicy supported negative_prompt to steer generation away from unwanted elements, that parameter is absent in the 3.0 Spicy schema. Workflows that relied on negative prompts will need to adjust by refining the positive prompt structure or handling unwanted results through post-filtering.

Pricing: Linear per-Second Billing

All three Wan 3.0 Spicy endpoints use the same pricing structure. Cost scales linearly with duration and resolution:

• 480p: $0.065 per second ($0.13 for 2s → $1.95 for 30s)
• 720p: $0.13 per second ($0.26 for 2s → $3.90 for 30s)
• 1080p: $0.26 per second ($0.52 for 2s → $7.80 for 30s)

The ref2v variant has one additional billing detail: billable duration = input video duration (up to 5 seconds) + output video duration. If you provide a 10-second reference video and generate a 15-second output, only the first 5 seconds of input are counted toward cost. The formula then applies the per-second rate above to the combined duration.

For t2v and i2v, pricing is straightforward duration × resolution. A 10-second 1080p clip costs $2.60 regardless of which endpoint generates it.

Start Generating with Wan 3.0 Spicy

Wan 3.0 Spicy is production-ready on Siray. Access all three endpoints — t2v, i2v, and ref2v — through the same API structure used across Siray's video models. Resolution scales up to 1080p, duration extends to 30 seconds in a single call, and the ref2v variant handles multi-modal reference input without external tooling.

Create your free Siray account and start generating uncensored video with Wan 3.0 Spicy.

Siray is committed to responsible AI use. All content generated through the platform must comply with applicable laws. Siray maintains zero tolerance for CSAM and any illegal content. Uncensored generation capabilities are provided exclusively for lawful adult content creation by and for consenting adults aged 18 and over.