Skip to main content

Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.Browse every model at this filtering level
bytedance/seedance-1.5-pro/text-to-video

Seedance 1.5 Pro text-to-video from ByteDance builds a whole shot from a written prompt and renders the soundtrack in the same pass. Clips run 4 to 12 seconds at 480p, 720p or 1080p, in six aspect ratios, with an optional camera lock.

From $0.011/s
Run model

Workbench

Input

Text to video
The only source of what appears on screen. Order it as subject, action, camera, light, sound, one idea per sentence, and restate the camera decision here when camera_fixed is true. Wording is not reviewed by us, but Seedance's own layer refuses part of it before rendering — kissing is the reliable trigger, while a whisper at the ear or foreheads touching passes. For a literal result, drive the beat from a first frame on image-to-video instead. at least 3 characters
Picks the framing from six fixed values, defaulting to 16:9. It changes cropping only — never length or sharpness. Take 9:16 for vertical placements, 16:9 for landscape, 21:9 for a widescreen feel. This generation has no adaptive option, so one value must be chosen.
affects price
Output resolution, 720p by default and 1080p at the top; nothing higher is offered on this endpoint. It sets a rate tier, so draft prompts at 480p where runs are cheap and quick, then raise it for the take you deliver.
Durationaffects price
Length of the finished clip in whole seconds, 4 to 12, defaulting to 5. This is the billed quantity — double the length, roughly double the cost. Keep one idea per shot and assemble longer pieces from several requests. no lower than 4 · no higher than 12
Generate audioaffects price
Whether a native soundtrack is rendered in sync with the picture. On by default and priced in a higher tier; switch it off when you plan to score, dub, or deliver silent footage.
Lock camera
Locks the camera, false by default. Left false, the model adds its own moves. Set to true for a static frame — product turntables, loopable plates, anything to be composited over a still background. Keep it false when you want a push, pan, or tilt, and describe that move in the prompt.
Seed
Integer from 0 to 2147483647 for the random start; leave it out and every run draws its own. Hold it fixed while you change one thing at a time to see what that change did — it keeps the starting noise the same, not the clip frame for frame. Available at 480p and 720p; a 1080p request that sets it is refused. No effect on price. no lower than 0 · no higher than 2147483647
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
Extend
Video editing
Video upscaling

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

One steady take, described in a sentence

Seedance 1.5 Pro's text-to-video endpoint turns a written prompt into one 4-12 second clip at 480p, 720p or 1080p, with a soundtrack rendered in the same pass. ByteDance published it as an audio-visual model, and its distinguishing field here is camera_fixed, a boolean that asks for a held frame instead of the movement the model would otherwise choose. There are no reference assets and no 4K tier on this generation.

Open the playground
One steady take, described in a sentence
Duration
4-12 whole seconds (duration_seconds), default 5
Resolution
480p / 720p / 1080p, default 720p
Aspect ratios
21:9, 16:9, 4:3, 1:1, 3:4, 9:16 — no adaptive option
Camera
camera_fixed asks for a held frame; false by default
Audio
generate_audio renders sound in the same pass, true by default
Reruns
Optional seed: integer from 0 to 2147483647 for controlled comparisons
Camera
The camera decision lives in a fieldcamera_fixed is part of the request rather than a phrase you write into the prompt and hope is honoured, so the decision survives every rewrite of the wording around it. Repeating it in the prompt reinforces it, and the first take is worth watching for drift before a clip goes into a composite.
Scope
Text and image in, no reference endpointThis family publishes text-to-video and image-to-video. Reference composition, where a cast or a location arrives as separate pictures, sits on Seedance 2.0, 2.0 Fast, 2.0 Mini and 2.5 instead. No Seedance family publishes video editing or extension either, so anything longer is cut together downstream.
Ceiling
1080p and 12 seconds are hard limitsNeither moves with a parameter. For 4K, the 2.0 generation is the one to route to: 2.0 Fast and 2.0 Mini stop at 720p, and 2.5 caps at 1080p like this family. Runtime is the other split, with the 2.0 line reaching 15 seconds and 2.5 reaching 30.

What the Seedance 1.5 Pro API does well

A camera that actually stays put

A camera that actually stays put

camera_fixed turns the camera into a setting instead of a suggestion. Left false, the model adds its own drift, push or slow orbit; set true, it holds the frame while the subject keeps moving inside it. Product turntables, talking heads and plates meant for compositing all depend on that stillness, so watch the first take for drift before an edit is built on top of it.

Synced audio by default

Synced audio by default

generate_audio is true unless you switch it off, so a request comes back as picture and sound together instead of a silent file waiting for a sound pass. ByteDance describes the model as generating voices and spatial effects that coordinate with the visuals, across a range of languages and dialects. Audio also selects a higher rate tier, which makes it a decision rather than a default to leave alone.

  • Six ratios cover the usual deliverables

    21:9, 16:9, 4:3, 1:1, 3:4 and 9:16 are the six values, 16:9 is the default, and there is no adaptive setting to fall back on. The ratio is part of the request, so a vertical cut is generated vertical rather than cropped out of a wide frame afterwards. Choose it for where the clip will be published before you spend a render on it.

  • A fair way to compare prompt variants

    Keep the prompt, seed, duration and resolution unchanged, then vary one setting at a time. On image-based tasks, reuse the same source images. Compare the staging and motion across a few takes to see whether the change helps.

  • A smaller contract, easier to reason about

    Text-to-video accepts prompt, duration_seconds, resolution, aspect_ratio, camera_fixed, generate_audio and optional seed. Set the camera and audio deliberately, and keep seed fixed when comparing prompt variants.

  • Check at 480p, deliver at 1080p

    Refine the shot at the lower resolution, then render the selected version at the delivery resolution with the same inputs and seed. Changing resolution starts a new generation, so review the final staging and motion before delivery.

Who Seedance 1.5 Pro is for

  • Lock the camera before writing the prompt

    Product and interview shots

    Lock the camera before writing the prompt

    Set camera_fixed to true first, then describe only what moves inside the frame: the product turning, the hands working, the subject speaking. The camera decision is made once and stays out of every later prompt revision, which is what keeps a run of takes cuttable against the same framing.

  • Aim straight at 9:16 or 1:1

    Short-form social cuts

    Aim straight at 9:16 or 1:1

    Choose the ratio that matches the destination feed before generating, because no adaptive value picks the crop for you and no second pass reframes a finished clip. Vertical work also holds up better when the prompt puts the subject in the upper half of the frame, clear of where a caption will sit.

  • Stay inside the family's own ceiling

    First-pass exploration under 12 seconds

    Stay inside the family's own ceiling

    Four to twelve seconds covers a single action, which is the shape this generation is built around. Start at 4 or 5 seconds to confirm the staging, then rerun the same wording at the length the cut actually needs. A brief that runs past twelve seconds belongs on the 2.0 line or on 2.5.

  • Generate the insert that was never shot

    Editors filling gaps in a cut

    Generate the insert that was never shot

    Nothing has to exist in front of a camera first, so coverage nobody filmed can be requested at the ratio the timeline already runs at, with audio on or off to match the track. Keep each insert to one action and one camera behaviour, then cut between several short takes instead of asking for a sequence.

How to call the Seedance 1.5 Pro API

  1. Decide the camera first

    camera_fixed shapes everything written after it, so set it before drafting the prompt. A held frame needs only the subject's movement described; a free camera needs a direction, a pace and a reason for the move as well.

  2. Pick the resolution for the job, not the ceiling

    480p is for checking wording and 1080p is for delivery, and both are separate requests: there is no upscale step that lifts a finished 480p clip. resolution also selects the rate tier, so make the call before submitting rather than after.

  3. Write a single beat

    Four to twelve seconds holds one action, so pick the beat before the wording: what changes between the first frame and the last. Give it one subject and one camera behaviour, and leave the chain of events to the edit.

  4. Set audio deliberately

    generate_audio is on unless you turn it off, and it moves the request into a higher rate tier. Leave it on when ambience or a spoken line ships with the clip, and switch it off for silent plates and loops headed into a sound edit.

Seedance 1.5 Pro worked examples

  • A static product hero shot

    camera_fixed true, 9 seconds, 1080p, 1:1 — built for a marketplace listing where any camera drift would look unintentional.

    A ceramic pour-over coffee dripper on a marble counter, steam rising slowly, warm morning light from the left, camera does not move

  • A short vertical talking-head beat

    camera_fixed true, 9:16, audio on — a clip that has to sit still while a person speaks, with room tone arriving in the same render.

    A person in a grey sweater speaking calmly to camera in a softly lit studio, subtle breathing motion only, steady frame, ambient room tone

  • A cinema-wide establishing beat with camera motion allowed

    camera_fixed false, 21:9, 12 seconds — the model handles a slow push instead of holding position, which is what you want when the move is part of the shot.

    Wide establishing shot of a foggy pine forest at dawn, camera slowly pushes forward between the trees, shafts of light cutting through the mist, birds calling in the distance

Seedance 1.5 Pro API questions

Decide three things before writing: whether the camera is locked, how long the beat needs to be, and whether sound ships with the clip. Then write the prompt for what changes on screen, since the model has no reference material to fall back on here. Run a 480p pass to read the wording back, and keep the 1080p request for the version you are ready to deliver.

All Seedance models10