Skip to main content

Seedance 2.5 Flash video edit is ByteDance's instruction-editing endpoint: send a 4–30 second clip and a text prompt, and it returns the clip re-rendered at 480p or 720p with the change applied while the performers, their movement and the camera stay as in the source. Turn generate_audio off to keep your clip's own soundtrack.

From $0.18/s

Workbench

Input

Video editing
590/20000

Reordering updates prompt references. Removing media also removes its references; review your prompt.

Open with the change, then list what must stay: people, clothing, props, movement and the camera move. Name reference images as @Image 1, @Image 2 and reference audio as @Audio 1, in upload order with a space before the number. at least 1 characters · at most 20000 characters
affects price1/1

Drop a video here, or browse · 1 file · Max 90 MB · MP4

  • 1
1 / 1 files
Link to the MP4 you want to edit, 4 to 30 seconds long; anything outside that range is refused before any charge. The result follows this clip's length, and the charge counts its seconds rounded up. at least 1 characters
1/10

Drop files or media links, paste, or browse · Up to 10 files · Max 10 MB · JPG · PNG · WEBP · GIF

  • @Image 1
1 / 10 files
Up to 10 optional images that show the target look, such as a garment, a prop or a colour board. Say in the prompt which @Image applies to which element; images do not change the charge. at least 1 items · at most 10 items
0/5

Drop an audio file here, or browse · Up to 5 files · Max 90 MB · MP3 · WAV

0 / 5 files
Up to 5 optional audio files, 30 seconds in total, that steer the soundtrack the model writes toward a voice or a mood. Pair them with generate_audio set to true; they add nothing to the charge. at least 1 items · at most 5 items
affects price
480p or 720p, default 720p. Settle the wording at 480p, then rerun the approved prompt at 720p; the choice sets the per-second rate applied to the source length.
Generate audio
true (default) writes a new soundtrack for the edited picture, and false keeps your clip's own sound. Our audio-off run returned a 44.1 kHz track matching the source, and the audio-on run came back 0.288 s shorter than its 4.0 s source.
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

Missing the model you need?

BYTEDANCE'S QUICK DRAFT TIER

Seedance 2.5 Flash API at a glance

Four Seedance 2.5 workflows at 480p and 720p, for trying a shot before you commit to a full render.

BytePlus speed claim
~2×
BytePlus says Flash gives results close to flagship Seedance 2.5 at nearly twice the generation speed. It is the publisher's figure, not a measurement of ours.
Resolution ceiling
720p
Flash renders 480p or 720p, with 720p as the default. BytePlus lists 1080p and 4K as not supported for this tier.
Clip length
4–30 s
Whole seconds, default 5, on text, image and reference requests. Video-edit takes a source clip of 4–30 seconds and follows its length.
Our 4 s 480p run times
84s–11m
Runs on 13 November 2026 finished between 84 seconds and about 11 minutes after submit, depending on the endpoint and the queue.

Results

What you can make with Seedance 2.5 Flash

  • Speed tier

    Flagship workflows at a lower ceiling

    Flash keeps the four Seedance 2.5 workflows that fit drafting (text, start and end frames, mixed references, clip editing), drops 1080p and 4K, and adds a video-extend built on its image-to-video mode. BytePlus presents the trade as results close to the flagship at nearly twice the speed. Our 4-second 480p renders took anywhere from 84 seconds to about 11 minutes, so measure speed on your own queue.

    Text-to-video at 720p for 5 seconds in 16:9, with the saxophone, the rain and his spoken line generated as sound.

  • Reference roles

    Give every upload a job in the prompt

    Name uploads by type and order with a space: @Image 1, @Video 1, @Audio 1. With two portraits as @Image 1 and @Image 2 in 16:9, both people appeared together in one elevator scene and kept their faces, hair and clothes. With one image, one clip and one voice file, the clip's street, yellow tram and tracking camera carried over, but the person stayed back to the camera and the face of @Image 1 barely showed.

    Reference-to-video at 720p in 16:9 for 5 seconds, with two portraits named @Image 1 and @Image 2 in the prompt.

  • Edit test

    A sunny street turned into a snowy evening

    Both riders, their bikes, the tram and the camera move matched the source frame by frame, and the weather and light changed while nothing else did. That was a 4.0 s clip at 480p with audio off, which also kept the original soundtrack.

  • Sound control

    One switch decides whose soundtrack you get

    generate_audio is on by default and produces speech, effects and ambience with the picture. In video-edit, switching it off keeps the source clip's own track (44.1 kHz in our test), while switching it on generates a new one that can follow @Audio 1. The switch has no effect on the price.

    Video edit at 720p: a 4.958-second sunny balcony clip turned into a rainy neon night, with her white sundress swapped for the red satin dress from @Image 1.

Before you run it

What to know before you run Seedance 2.5 Flash

All modes
Flash stops at 720p: BytePlus lists 1080p and 4K as not supported for this tier, and there is no draft flag or upscale step. Flash video-extend continues a clip from its last frame at 480p or 720p; send the brief to Seedance 2.5 when you need 1080p, 4K or an extension that reads the clip's final seconds.
All modes
Each request is one render with a length fixed between 4 and 30 whole seconds, so longer sequences are assembled from several clips. In our 480p, 4-second runs the time from submit to finish ranged from 84 seconds to about 11 minutes, and longer or 720p clips take longer, so poll the task or pass callBackUrl instead of holding a connection open.
Reference to video
Reference clips total up to 30 seconds, count as at least 2 seconds each and need about 409,600 pixels or more: 854×480 passes, while a 640×360 clip was rejected at submit. Upload faces, voices, products and footage you have the right to use, and do not present a generated person as someone real.
Video editing
The source clip must run 4 to 30 seconds; anything outside that range is rejected at submit. The output follows the source length: a 4.0 s source returned 4.000 s with audio off and 3.712 s with audio on, and billing counts source seconds either way.
All modes
The model can decline or tone down an individual request, so compare each result with the brief; failed generations are not charged. Finished files stay available for 14 days and are then deleted, so download what you want to keep.

Use cases

What to build with the Seedance 2.5 Flash API

  • Trying a shot plan before a final render

    Send a 4-second, 480p, 16:9 prompt that ends with an Audio: line, such as a rain-soaked night crosswalk. Our run returned 854×480 at 24 fps, 4.064 s long, with an AAC stereo track of rain and laughter. Change one phrase and rerun to see what moved.

  • Restyling footage you already shot

    Upload a 4–30 second clip, describe the change, then list what must stay. A sunny street with two cyclists and a tram became a snowy evening in our test, and with generate_audio switched off the original soundtrack came back unchanged.

  • Character and product mockups

    Attach a portrait and a product photo as @Image 1 and @Image 2 and ask for 9:16. Our two-image test returned 480×854 with the man lifting a glass dropper bottle toward the lens, so both references showed up. Use the result to judge the idea before building a polished version.

  • Looping a portrait animation

    Give the same photo as image_url and last_image_url and describe the motion between them. Our potter portrait returned a 562×748 clip of 4.064 s whose opening and closing frames both matched the source pose.

  • Chaining one shot into the next

    Set return_last_frame to true and the MP4 arrives with a JPEG of its closing frame (854×480 at 480p 16:9 in our test). Pass that JPEG as image_url of the next image-to-video request so the following shot begins where the last one stopped.

Prompts

Seedance 2.5 Flash prompt examples you can copy

  • Text to video

    Lighthouse stair at dawn

    A woman in her fifties with silver braids and a mustard wool coat climbs the spiral stairs of a coastal lighthouse, one hand trailing the iron rail. She reaches the lamp room, wipes fog from the glass with her sleeve and looks out at the grey sea. Slow vertical crane up alongside the stairwell, cool window light mixed with the warm glow of an oil lantern. Audio: boots ringing on metal steps, wind in the tower, a distant foghorn, no music.

    Why it works. One climb, one reveal and one camera path keep the beat count low, and the Audio line names three sources the model can place in time. For an opening test, render 4 seconds at 480p.

    Try this prompt
  • Text to video

    Night-market wok in 9:16

    A street cook in his forties tosses noodles in a roaring wok at a night-market stall, flames licking up, then slides a bowl across the counter to a waiting customer who bows slightly. Handheld medium shot, steam backlit by strings of bulbs, shallow focus on his hands. Audio: sizzling oil, the scrape of a metal spatula, a murmuring crowd, no music.

    Why it works. Ask for a 9:16 frame to get a vertical clip. A single repeated motion (the toss) and a clear end beat (the bowl) give the model a start and a finish inside four to six seconds.

    Try this prompt
  • Image to video

    Violinist draws one note

    From the opening pose, the violinist lifts her bow, draws one long note and turns her eyes toward the window as the camera drifts slowly closer. Late-afternoon light slides across the wooden floor and her dress. Audio: one sustained violin note with room reverb, a faint chair creak, no other music.

    Why it works. The start image already fixes who and where, so the prompt spends its words on motion, camera and sound. The output takes the shape of the start image.

    Try this prompt
  • Image to video

    Bakery shutter opens

    The start frame shows a bakery with its shutter down at blue dawn; the end frame shows the same doorway open with the lights on. The metal shutter rolls up, warm light spills onto the wet pavement and the baker, a man in his sixties in a flour-dusted apron, steps out to turn the sign. Locked-off wide shot, blue light shifting to amber. Audio: a rattling shutter, a bell over the door, waking birds, no music.

    Why it works. Supply the open doorway as the end frame and describe the path between the two pictures. Keep the framing of both images identical so the move reads as one continuous shot.

    Try this prompt
  • Reference to video

    Courier on the quay

    The courier in the green rain jacket from @Image 1 rides along a harbor quay with the same low tracking move as @Video 1, slows at a mooring post and glances at the lens with a short smile. He says one line in the voice from @Audio 1: "Parcel for the harbor office." Overcast light, wet cobblestones. Audio: tires on stone, a gull overhead, his spoken line.

    Why it works. Each reference has one role: the look from @Image 1, the camera from @Video 1 and the voice from @Audio 1. Reference clips need about 409,600 pixels, so use 854×480 or larger.

    Try this prompt
  • Reference to video

    Teapot pour at a window table

    The woman from @Image 1 sits at a sunlit cafe table by a rain-streaked window, lifts the ceramic teapot from @Image 2, pours a thin stream into a cup and breathes in the steam with a content smile. Static medium shot, soft window light with reflections on the glazed surface. Audio: water pouring, porcelain clinking, a quiet room, no music.

    Why it works. A person image plus an object image mirrors our two-image test, where both references appeared in the clip. Name what the object does in the scene, not just what it looks like.

    Try this prompt
  • Video editing

    Beach walk to stormy dusk

    Turn the sunny beach into a stormy dusk with low grey clouds, wet sand and a light drizzle. Keep both walkers, their coats, their pace, the dog and the slow tracking camera exactly as in the source video. Audio: gusting wind, far-off surf, soft drizzle on fabric.

    Why it works. Lead with the change, then list what must stay, as in our tram-street edit. Leave sound on to get the new soundtrack described here, or turn it off to keep the clip's own sound.

    Try this prompt

Pricing

Seedance 2.5 Flash API pricing

A 5-second clip starts at $0.575 at 480p and $1.2 at 720p.

Get started

How to use the Seedance 2.5 Flash API in three steps

  1. Step 1

    Choose the endpoint for your input

    Use text-to-video for a prompt alone, image-to-video for a start frame, reference-to-video for images, clips or a voice, video-edit for an existing 4–30 second clip, and video-extend to add 4–30 seconds to a clip of up to 60 seconds. All five share the 480p/720p range and the generate_audio switch.

  2. Step 2

    Send the createTask request

    POST to https://api.spicyapi.ai/api/v1/jobs/createTask with the header Authorization: Bearer YOUR_SPICY_API_KEY and the body {"model": "bytedance/seedance-2.5-flash/text-to-video", "input": {"prompt": "A lighthouse keeper climbs a spiral stair at dawn and wipes fog from the lamp glass. Audio: wind, boots on iron.", "duration_seconds": 4, "resolution": "480p"}}.

  3. Step 3

    Poll for the finished clip

    The response gives a taskId in state queued. Call GET /api/v1/jobs/recordInfo?taskId=YOUR_TASK_ID, or pass a callBackUrl, until the state is succeeded, then read the MP4 address from output.assets. Field details are in the task guide at docs.spicyapi.ai/docs/tasks.

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

FAQ

Seedance 2.5 Flash API FAQ

About Seedance 2.5 Flash

Seedance 2.5 Flash suits the stage where a shot is still a question. Write the brief, run it at 480p for four seconds, and change one clause between runs so every difference has a cause. Our runs on 13 November 2026 finished between 84 seconds and about 11 minutes, so batch your tests and use polling or a callback instead of waiting on one request. When timing and framing hold, keep the 720p result, or send the same prompt and references to the Seedance 2.5 page when you need 1080p or 4K. To build a longer sequence at 480p or 720p, continue the take with Flash video-extend.

Related models and next tasks