Seedance 2.5 Spicy NSFW Prompt Tutorial for t2v/i2v

bytedance/seedance-2.5-t2v-spicy and bytedance/seedance-2.5-i2v-spicy, both available through Siray's API, are ByteDance's uncensored variants of Seedance 2.5 — live since 2026-08-14. This tutorial covers prompting for these two endpoints specifically. A third variant, ref2v, also exists on this model but is out of scope here; it takes reference images, videos, and audio and deserves its own dedicated walkthrough rather than a mention in passing.

Key Takeaways

• Seedance 2.5 Spicy runs size at 480p, 720p, or 1080p and duration at 4-30 whole seconds — the 30-second ceiling ties with Wan 3.0 Spicy for the longest single-call NSFW video generation in Siray's pool, not an exclusive record.
• Pricing scales with resolution: 480p $0.140/s, 720p $0.302/s, 1080p $0.739/s — no discount currently applied.
• audio_enable is a real field on this endpoint; decide whether you want native audio before you set it.

Overview: What t2v and i2v Give You Here

t2v generates from text alone; i2v animates a starting image. Both endpoints share the same size, duration, and audio_enable fields, and both are priced identically at a given resolution — the difference is purely the input type. Every technique below works with positive, descriptive prompting: build the prompt around what you want in frame, not around exclusions.

Element 1: Subject and Action

This is the core of the prompt — describe who's in frame and what they're doing, in plain, specific language. Vague verbs ("moving") produce vague motion; specific ones ("arching her back slowly," "walking toward the camera with deliberate steps") give the model something concrete to render.Mini example: "A woman kneels on a bed, slowly leaning forward, hair falling across her shoulder."

Element 2: Scene, Setting, and Lighting

Name the room or environment and the light source — this anchors the whole shot and avoids a flat, generic background. "Warm lamp light" and "cool moonlight through a window" produce visibly different results from the same subject description.Mini example: "in a dimly lit bedroom, warm lamp light from a bedside table"

Element 3: Camera Framing and Motion

Since this is video, describe the shot type and any camera movement, not just the subject. "Close-up" vs. "wide shot," "static camera" vs. "slow push-in" — both change pacing and mood independent of what the subject is doing.Mini example: "close-up shot, slow push-in, shallow depth of field"

Element 4: Duration and Its Tradeoffs

duration runs 4-30 whole seconds. Longer clips cost proportionally more at every resolution, and a prompt written for a 5-second beat won't automatically fill 20 seconds with new content — it'll often just extend the same motion. For a single continuous action, keep duration matched to how long that action actually takes; don't default to the max just because it's available.

Element 5: Resolution and Price

size is 480p, 720p, or 1080p, and price tracks resolution: 480p $0.140/s, 720p $0.302/s, 1080p $0.739/s, with no discount currently applied. A 10-second clip costs $1.40 at 480p versus $7.39 at 1080p — decide how much resolution the final use actually needs before defaulting to 1080p for every test generation.

Element 6: audio_enable

This boolean controls whether the model generates native audio alongside the video. Turn it on for a scene where ambient sound adds to the effect (breathing, movement sounds); leave it off for a silent clip you'll score separately, since generating audio you'll discard doesn't save anything on cost.

Putting It All Together

t2v prompt: "A woman kneels on a bed, slowly leaning forward, hair falling across her shoulder, in a dimly lit bedroom with warm lamp light from a bedside table. Close-up shot, slow push-in, shallow depth of field." — duration: 8, size: "720p", audio_enable: true.

i2v prompt (starting from a still image): "She turns toward the camera and leans back against the headboard, warm lamp light steady across her face. Slow zoom out, static frame otherwise." — duration: 6, size: "1080p", audio_enable: false.

Every element above is present in both: subject/action, setting/lighting, camera framing, and an explicit duration/resolution/audio choice — not left to defaults.

Developer Notes

Both endpoints return through the same normalized response format Siray uses across its catalog, so a t2v prompt built for testing can be adapted to i2v (and vice versa) by swapping the input type, not rebuilding the request from scratch. For the ref2v variant of this model — reference-image, video, and audio inputs — that's a separate workflow this tutorial deliberately doesn't cover. For the plain (non-Spicy) Seedance 2.5 workflow, see Seedance 2.5: A Brand Explainer and Video Workflow.

Ready to try it? Create your free Siray account and start generating with Seedance 2.5 Spicy t2v and i2v.

A Note on Compliance

Siray's own NSFW and uncensored models generate legal, consensual adult content only — CSAM and any illegal content are refused with zero tolerance. "Spicy" means no moderation beyond what the law requires, not that anything goes. Seedance 2.5 Spicy is a third-party ByteDance model available through Siray's API, not a Siray-hosted model, and the same standard applies to everything generated with it.