Seedance 2.5 Video to Video API

bytedance/seedance-v2.5/video-to-video

Elinizdeki görüntüyü yeni bir videoya dönüştürün — klibinizi yeniden tarzlayın, sahneleyin, çerçeveleyin ya da uzatın; prompt’ta andığınız en fazla 10 kaynak videoyla yönlendirin.

düzenlemeuzatma

1M token başına $9.6 çıktı

OpenAPI

Input

Prompt

SOURCE AND CONTINUATION
Extend @video1. This is a direct continuation that begins at the exact frame where @video1 ends — the same ride, the same day, the next stretch of desert. Every shot below is new material that happens after the source footage, not a retelling of it.
Preserve exactly from @video1: the rider's identity, helmet and visor, black riding suit, gloves and boots; the hoverboard and its cyan underglow; the desert world, dune shapes and haze; the low golden sun and its direction; the colour grade; and the music, which continues unbroken in the same key, tempo and arrangement with no restart.
CHARACTERS
The rider: same man from @video1, full dark helmet with a clear visor and two white light strips, black armored riding suit, gloves, boots. Identical in every shot. He rides the same hoverboard, lit underneath by a continuous cyan strip.
LOCATION MAP
Foreground: open sand, wind-carved ripples, loose grit lifting off the crests.
Midground: the ride path — dunes flattening into a cracked salt pan that runs into a post-apocalyptic ruin field of collapsed concrete slabs, leaning steel columns, half-buried vehicle hulls and broken roadway sections.
Background: a low sandstone cliff wall on the far side of the ruins, and beyond it open sky with the sun sitting low on the horizon.
The rider travels camera-left to camera-right in every lateral view. The sun stays low and ahead of him, camera-right, throughout.
FIRST FRAME / BLOCKING
Full-frame lateral profile close-up of the rider, 18°, camera tracking alongside him at his own speed, his helmet filling the right third of frame in sharp profile, dunes streaking past behind him, cyan board glow spilling up onto his boots from below.
FORMAT MODE
Timed multishot with HARD CUTS only at 4.0s, 8.0s, 12.0s, 15.0s, 19.0s, 23.5s and 27.0s. The camera cuts only at these points. Real-time playback everywhere except 19.0–23.5s, which is slow motion; the speed change happens only on those two hard cuts.
STORYBOARD
SHOT 1 (0.0–4.0s)
Lateral profile close-up, 18°, camera tracking beside him at 60 km/h, level with his shoulder. He rides steady, head turning a few degrees forward as the ruins come into his sightline, then his shoulders drop slightly and he leans in. Sand streams off the board's trailing edge behind him. Low sun rakes across the front of his visor from camera-right, leaving the far side of his helmet in deep shade.
SHOT 2 (4.0–8.0s)
HARD CUT to an extreme wide, 84°, camera high and static as he crosses the frame camera-left to camera-right at 60 km/h, a small figure with a cyan thread of light beneath him. Ahead of him the ruin field opens up — collapsed slabs and leaning steel standing as tall as six humans, throwing long shadows across the salt pan.
SHOT 3 (8.0–12.0s)
HARD CUT to a low tracking shot, 47°, camera 30 cm off the ground running with him at 70 km/h as he enters the ruins. He weaves between two leaning columns, board tilting hard onto its edge, a sheet of sand and grit fanning off the tail. Broken concrete passes close on both sides. Cyan underglow flares against the slab faces as he passes them.
SHOT 4 (12.0–15.0s)
HARD CUT to a medium shot from the front quarter, 29°, camera drifting backward ahead of him at 70 km/h. He drops into a deep drift, both arms out for balance, the board sliding sideways through a curve of debris, a long curtain of dust thrown out behind him and catching the low sun. His head stays locked on his line through the turn.
SHOT 5 (15.0–19.0s)
HARD CUT to a low rear three-quarter shot, 63°, camera accelerating with him from 70 to 110 km/h. He crouches, drives his weight forward, and the board's cyan strip brightens as it surges. Ahead of him a broken roadway section rises out of the sand at a shallow angle, its torn edge lifted clear of the ground. He hits it square and launches.
SHOT 6 (19.0–23.5s)
HARD CUT into SLOW MOTION. Wide side view, 47°, camera craning slowly upward with him against the low sun. He is airborne at the apex, board still under his feet, one hand gripping the front edge, knees folded. A full curtain of sand hangs suspended around and beneath him, every grain lit gold from camera-right and drifting rather than falling. The cyan glow underneath him rakes across the suspended sand. The whole shot stays in slow motion start to finish.
SHOT 7 (23.5–27.0s)
HARD CUT back to real time. Low tracking shot, 47°, camera behind and beside him. He lands hard, board slapping the sand and compressing, a burst of grit blowing outward from the impact, and he rides straight out of it at 110 km/h across the last open stretch toward the sandstone cliffs. The ruins fall away behind him.
SHOT 8 (27.0–30.0s)
HARD CUT to a lateral profile medium shot, 29°, camera settling alongside him and coming to rest as he does. He carves the board hard sideways and brakes, throwing a low wide sheet of sand forward and to camera-right, and stops at the cliff edge. He straightens up, squares his shoulders, and turns his helmet to face the sun sitting on the horizon. Held final frame: the rider in profile at the cliff edge, sand still settling around the board, the cyan glow steady beneath him, the sun low ahead of him.
PERFORMANCE
Everything reads through the body, since his face is behind the visor: weight shifting into each turn, arms counterbalancing the drift, the deep crouch before the launch, the folded stillness at the apex, the absorbed compression of the landing, and the settled straightening at the very end. Restrained and controlled throughout — a rider who knows this ground.
PHYSICS
The board carries real mass and inertia. It leans into turns and rebounds on landing. Sand behaves as loose dry grain: it fans off the edge under load, hangs in a curtain in slow motion, blows outward in a burst on impact, and settles slowly at the end. Dust catches and scatters the low sun. Wind drags loose grit off the dune crests continuously. Debris in the ruins stays put; only sand and dust move.
LIGHTING
Fixed 3200K golden-hour sun, low on the horizon ahead of the rider and camera-right, unchanged in every shot. Hard warm key on the front of his suit and visor, deep shade on the far side, long shadows running away camera-left across the sand and the ruin floor. Airborne dust glows where it crosses the sun. Cyan board light acts as a cold practical from below, catching his boots, the sand under him and nearby slab faces.
AUDIO
The music from @video1 continues unbroken through the whole extension — same track, same key, same tempo, no restart and no new song. Under it: the low hum of the board, wind across open sand, grit spraying off the edge, concrete and steel echoing close in the ruins.
At 19.0s the mix thins for the slow-motion shot — the music drops to a single sustained low note, and the suspended sand and the board's hum sit at the front.
At 23.5s the full arrangement returns hard on the landing impact.
In the final shot the music resolves and holds under open desert wind. No dialogue anywhere.
OUTPUT SETTINGS
16:9, real-time in every segment except 19.0–23.5s, which is slow motion.
POSITIVE LOCKS
This is a continuation of @video1 — all action takes place after the source footage ends.
The rider, helmet, visor light strips, suit and the hoverboard's cyan underglow stay identical to @video1 in every shot.
The rider travels camera-left to camera-right in every lateral view, and the sun stays low and camera-right in every shot.
One continuous piece of music runs from the first frame to the last, carried over from @video1.
The board keeps complete geometry and stays under his feet through the jump and landing.
Cuts land only at the stated seconds, and the speed change to slow motion happens only at 19.0s and back at 23.5s.
The video ends on the held profile of the rider stopped at the cliff edge, facing the low sun.

Describe the result you want and what to take from each reference. Address the reference assets positionally as @Video1..@Video10, @Image1..@Image30 and @Audio1..@Audio10, numbered by their order in reference_videos, reference_images and reference_audios: with one source clip and one product photo, 'Match the camera movement and pacing of @Video1, but restage it on a sunlit desert roadside at midday; the rider pulls up on the motorcycle from @Image1' borrows the motion from the clip and the subject from the image. A reference the prompt never names is usually ignored, so name every one you send.

Reference Videos

Drag & drop videos here

or click to browse

1 to 10 source clips - the footage this generation is built from, and where its motion, camera movement and pacing come from. The first entry is @Video1 in the prompt, the second @Video2, and so on. MP4 or MOV (H.264/H.265 video, AAC or MP3 audio), 2-30 seconds per file and no more than 30 seconds COMBINED across all files, 300-6000 px per side, aspect ratio between 0.4 and 2.5, 24-60 fps, under 200 MB each. Billing counts the duration of everything you send here as well as the length of the clip you get back, so a 30-second source costs about as much again as a 30-second output - trim each clip to the part that matters.

Reference Images

Drag & drop images here

or click to browse

Optional: up to 30 reference images that steer identity, wardrobe, product, location or style alongside the source footage. Addressed as @Image1, @Image2 and so on, numbered by their order in this array. JPEG, PNG, WebP, BMP, TIFF or GIF; 300-6000 px on a side, aspect ratio between 1:2.5 and 2.5:1, under 30 MB each. A still image has no duration, so reference images add nothing to the price.

resolution

Output resolution of the clip. 720p meters more tokens per second of video than 480p, so it costs more.

aspect_ratio

Frame shape of the generated clip. The source footage does not dictate the framing on this variant, so pick the shape you want - a 1280x720 (16:9) source asked for 9:16 came back as a 480x854 portrait clip. Use adaptive to keep the source's own shape instead.

Duration

Length of the generated clip in seconds, from 4 to 30. Leave at -1 (the default) to let the model choose a whole-second length close to the source footage. Billing counts this alongside the duration of the source clips you send.

Additional Settings

Customize your input with more control.

Generate Audio

Generate a synchronized soundtrack (dialogue, ambience and sound effects) together with the picture. Set false for a silent clip; the price is the same either way.

Reference Audios

Drag & drop audio files here

or click to browse

Experimental: up to 10 short audio references for the generated soundtrack, addressed as @Audio1, @Audio2 and so on. Each clip 2-30 seconds, 30 seconds combined at most; WAV or MP3, under 15 MB each. The clips are accepted and add nothing to the price, but their effect on the finished soundtrack has not been verified - treat the field as experimental. They cannot be sent on their own: at least one reference video is always required.

You need to be logged in to run this model and view results.

Output

{
  "error": "",
  "inferenceTime": 0,
  "output": "https://media.modelrunner.ai/aFCT7ZGsmPGSqbLe-desert-hoverboard-ruin-run-output.webm",
  "input": {
    "prompt": "SOURCE AND CONTINUATION\nExtend @video1. This is a direct continuation that begins at the exact frame where @video1 ends — the same ride, the same day, the next stretch of desert. Every shot below is new material that happens after the source footage, not a retelling of it.\nPreserve exactly from @video1: the rider's identity, helmet and visor, black riding suit, gloves and boots; the hoverboard and its cyan underglow; the desert world, dune shapes and haze; the low golden sun and its direction; the colour grade; and the music, which continues unbroken in the same key, tempo and arrangement with no restart.\nCHARACTERS\nThe rider: same man from @video1, full dark helmet with a clear visor and two white light strips, black armored riding suit, gloves, boots. Identical in every shot. He rides the same hoverboard, lit underneath by a continuous cyan strip.\nLOCATION MAP\nForeground: open sand, wind-carved ripples, loose grit lifting off the crests.\nMidground: the ride path — dunes flattening into a cracked salt pan that runs into a post-apocalyptic ruin field of collapsed concrete slabs, leaning steel columns, half-buried vehicle hulls and broken roadway sections.\nBackground: a low sandstone cliff wall on the far side of the ruins, and beyond it open sky with the sun sitting low on the horizon.\nThe rider travels camera-left to camera-right in every lateral view. The sun stays low and ahead of him, camera-right, throughout.\nFIRST FRAME / BLOCKING\nFull-frame lateral profile close-up of the rider, 18°, camera tracking alongside him at his own speed, his helmet filling the right third of frame in sharp profile, dunes streaking past behind him, cyan board glow spilling up onto his boots from below.\nFORMAT MODE\nTimed multishot with HARD CUTS only at 4.0s, 8.0s, 12.0s, 15.0s, 19.0s, 23.5s and 27.0s. The camera cuts only at these points. Real-time playback everywhere except 19.0–23.5s, which is slow motion; the speed change happens only on those two hard cuts.\nSTORYBOARD\nSHOT 1 (0.0–4.0s)\nLateral profile close-up, 18°, camera tracking beside him at 60 km/h, level with his shoulder. He rides steady, head turning a few degrees forward as the ruins come into his sightline, then his shoulders drop slightly and he leans in. Sand streams off the board's trailing edge behind him. Low sun rakes across the front of his visor from camera-right, leaving the far side of his helmet in deep shade.\nSHOT 2 (4.0–8.0s)\nHARD CUT to an extreme wide, 84°, camera high and static as he crosses the frame camera-left to camera-right at 60 km/h, a small figure with a cyan thread of light beneath him. Ahead of him the ruin field opens up — collapsed slabs and leaning steel standing as tall as six humans, throwing long shadows across the salt pan.\nSHOT 3 (8.0–12.0s)\nHARD CUT to a low tracking shot, 47°, camera 30 cm off the ground running with him at 70 km/h as he enters the ruins. He weaves between two leaning columns, board tilting hard onto its edge, a sheet of sand and grit fanning off the tail. Broken concrete passes close on both sides. Cyan underglow flares against the slab faces as he passes them.\nSHOT 4 (12.0–15.0s)\nHARD CUT to a medium shot from the front quarter, 29°, camera drifting backward ahead of him at 70 km/h. He drops into a deep drift, both arms out for balance, the board sliding sideways through a curve of debris, a long curtain of dust thrown out behind him and catching the low sun. His head stays locked on his line through the turn.\nSHOT 5 (15.0–19.0s)\nHARD CUT to a low rear three-quarter shot, 63°, camera accelerating with him from 70 to 110 km/h. He crouches, drives his weight forward, and the board's cyan strip brightens as it surges. Ahead of him a broken roadway section rises out of the sand at a shallow angle, its torn edge lifted clear of the ground. He hits it square and launches.\nSHOT 6 (19.0–23.5s)\nHARD CUT into SLOW MOTION. Wide side view, 47°, camera craning slowly upward with him against the low sun. He is airborne at the apex, board still under his feet, one hand gripping the front edge, knees folded. A full curtain of sand hangs suspended around and beneath him, every grain lit gold from camera-right and drifting rather than falling. The cyan glow underneath him rakes across the suspended sand. The whole shot stays in slow motion start to finish.\nSHOT 7 (23.5–27.0s)\nHARD CUT back to real time. Low tracking shot, 47°, camera behind and beside him. He lands hard, board slapping the sand and compressing, a burst of grit blowing outward from the impact, and he rides straight out of it at 110 km/h across the last open stretch toward the sandstone cliffs. The ruins fall away behind him.\nSHOT 8 (27.0–30.0s)\nHARD CUT to a lateral profile medium shot, 29°, camera settling alongside him and coming to rest as he does. He carves the board hard sideways and brakes, throwing a low wide sheet of sand forward and to camera-right, and stops at the cliff edge. He straightens up, squares his shoulders, and turns his helmet to face the sun sitting on the horizon. Held final frame: the rider in profile at the cliff edge, sand still settling around the board, the cyan glow steady beneath him, the sun low ahead of him.\nPERFORMANCE\nEverything reads through the body, since his face is behind the visor: weight shifting into each turn, arms counterbalancing the drift, the deep crouch before the launch, the folded stillness at the apex, the absorbed compression of the landing, and the settled straightening at the very end. Restrained and controlled throughout — a rider who knows this ground.\nPHYSICS\nThe board carries real mass and inertia. It leans into turns and rebounds on landing. Sand behaves as loose dry grain: it fans off the edge under load, hangs in a curtain in slow motion, blows outward in a burst on impact, and settles slowly at the end. Dust catches and scatters the low sun. Wind drags loose grit off the dune crests continuously. Debris in the ruins stays put; only sand and dust move.\nLIGHTING\nFixed 3200K golden-hour sun, low on the horizon ahead of the rider and camera-right, unchanged in every shot. Hard warm key on the front of his suit and visor, deep shade on the far side, long shadows running away camera-left across the sand and the ruin floor. Airborne dust glows where it crosses the sun. Cyan board light acts as a cold practical from below, catching his boots, the sand under him and nearby slab faces.\nAUDIO\nThe music from @video1 continues unbroken through the whole extension — same track, same key, same tempo, no restart and no new song. Under it: the low hum of the board, wind across open sand, grit spraying off the edge, concrete and steel echoing close in the ruins.\nAt 19.0s the mix thins for the slow-motion shot — the music drops to a single sustained low note, and the suspended sand and the board's hum sit at the front.\nAt 23.5s the full arrangement returns hard on the landing impact.\nIn the final shot the music resolves and holds under open desert wind. No dialogue anywhere.\nOUTPUT SETTINGS\n16:9, real-time in every segment except 19.0–23.5s, which is slow motion.\nPOSITIVE LOCKS\nThis is a continuation of @video1 — all action takes place after the source footage ends.\nThe rider, helmet, visor light strips, suit and the hoverboard's cyan underglow stay identical to @video1 in every shot.\nThe rider travels camera-left to camera-right in every lateral view, and the sun stays low and camera-right in every shot.\nOne continuous piece of music runs from the first frame to the last, carried over from @video1.\nThe board keeps complete geometry and stays under his feet through the jump and landing.\nCuts land only at the stated seconds, and the speed change to slow motion happens only at 19.0s and back at 23.5s.\nThe video ends on the held profile of the rider stopped at the cliff edge, facing the low sun.",
    "duration": 30,
    "resolution": "480p",
    "aspect_ratio": "16:9",
    "generate_audio": true,
    "reference_videos": [
      "https://media.modelrunner.ai/8pxYmpUxfJcx19Jp-desert-hoverboard-ruin-run-source1.webm"
    ]
  },
  "logs": ""
}

Logs (1 lines)

Örnekler

Fiyatlandırma

Videonun saniyesi üzerinden faturalandırılır — gönderdiğiniz kaynak klip de aldığınız klip ile birlikte ölçülür.

Tipik çalıştırma

≈$2.07

for a 5s 720p clip

Çıktı token’ları $9.6 1M başına

Input + output duration are billed · smallest run $0.56

Bir çalıştırma token kullanımı bildirmezse sabit $2.07 faturalandırılır.

Seedance 2.5 Video to Video API

Seedance 2.5 Video to Video is a video-to-video AI model by bytedance. On ModelRunner it runs through a REST API or via MCP from any AI assistant, at about $2.07 for a 5s 720p clip.

POST https://queue.modelrunner.run/bytedance/seedance-v2.5/video-to-video

cURL

# Submit a request to the queue. Input fields go at the top level of the
# body. The optional reserved "metadata" object holds your own flat string
# tags — stored on the request, never sent to the model; filter later with
# GET https://queue.modelrunner.run/requests?metadata=<url-encoded JSON>.
curl -X POST https://queue.modelrunner.run/bytedance/seedance-v2.5/video-to-video \
  -H "Authorization: Key $MRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "SOURCE AND CONTINUATION\nExtend @video1. This is a direct continuation that begins at the exact frame where @video1 en…",
    "duration": 30,
    "resolution": "480p",
    "aspect_ratio": "16:9",
    "generate_audio": true,
    "reference_videos": [
      "https://media.modelrunner.ai/8pxYmpUxfJcx19Jp-desert-hoverboard-ruin-run-source1.webm"
    ],
    "metadata": {
      "project": "my-project"
    }
  }'
# → { "request_id": "...", "status_url": "...", "response_url": "..." }

# Poll status_url until "COMPLETED", then fetch the result
curl "https://queue.modelrunner.run/bytedance/seedance-v2.5/video-to-video/requests/$REQUEST_ID/status" \
  -H "Authorization: Key $MRUN_API_KEY"
curl "https://queue.modelrunner.run/bytedance/seedance-v2.5/video-to-video/requests/$REQUEST_ID" \
  -H "Authorization: Key $MRUN_API_KEY"

JavaScript

import { modelrunner } from "@modelrunner/client";

const result = await modelrunner.subscribe("bytedance/seedance-v2.5/video-to-video", {
  input: {
    "prompt": "SOURCE AND CONTINUATION\nExtend @video1. This is a direct continuation that begins at the exact frame where @video1 en…",
    "duration": 30,
    "resolution": "480p",
    "aspect_ratio": "16:9",
    "generate_audio": true,
    "reference_videos": [
      "https://media.modelrunner.ai/8pxYmpUxfJcx19Jp-desert-hoverboard-ruin-run-source1.webm"
    ]
  },
});
console.log(result);

Python

import os
import requests

headers = {"Authorization": f"Key {os.environ['MRUN_API_KEY']}"}

submitted = requests.post(
    "https://queue.modelrunner.run/bytedance/seedance-v2.5/video-to-video",
    headers=headers,
    json={
      "prompt": "SOURCE AND CONTINUATION\nExtend @video1. This is a direct continuation that begins at the exact frame where @video1 en…",
      "duration": 30,
      "resolution": "480p",
      "aspect_ratio": "16:9",
      "generate_audio": true,
      "reference_videos": [
        "https://media.modelrunner.ai/8pxYmpUxfJcx19Jp-desert-hoverboard-ruin-run-source1.webm"
      ]
    },
).json()

# Poll submitted["status_url"] until "COMPLETED", then:
result = requests.get(submitted["response_url"], headers=headers).json()

Input parameters

Input parameters of Seedance 2.5 Video to Video
Name	Type	Required	Description
prompt	string	yes	Describe the result you want and what to take from each reference. Address the reference assets positionally as @Video1..@Video10, @Image1..@Image30 and @Audio1..@Audio10, numbered by their order in reference_videos, reference_images and reference_audios: with one source clip and one product photo, 'Match the camera movement and pacing of @Video1, but restage it on a sunlit desert roadside at midday; the rider pulls up on the motorcycle from @Image1' borrows the motion from the clip and the subject from the image. A reference the prompt never names is usually ignored, so name every one you send.
reference_videos	array	yes	1 to 10 source clips - the footage this generation is built from, and where its motion, camera movement and pacing come from. The first entry is @Video1 in the prompt, the second @Video2, and so on. MP4 or MOV (H.264/H.265 video, AAC or MP3 audio), 2-30 seconds per file and no more than 30 seconds COMBINED across all files, 300-6000 px per side, aspect ratio between 0.4 and 2.5, 24-60 fps, under 200 MB each. Billing counts the duration of everything you send here as well as the length of the clip you get back, so a 30-second source costs about as much again as a 30-second output - trim each clip to the part that matters.
reference_images	array	no	Optional: up to 30 reference images that steer identity, wardrobe, product, location or style alongside the source footage. Addressed as @Image1, @Image2 and so on, numbered by their order in this array. JPEG, PNG, WebP, BMP, TIFF or GIF; 300-6000 px on a side, aspect ratio between 1:2.5 and 2.5:1, under 30 MB each. A still image has no duration, so reference images add nothing to the price.
resolution	enum	no	Output resolution of the clip. 720p meters more tokens per second of video than 480p, so it costs more. One of: `480p`, `720p`. Default: "720p".
aspect_ratio	enum	no	Frame shape of the generated clip. The source footage does not dictate the framing on this variant, so pick the shape you want - a 1280x720 (16:9) source asked for 9:16 came back as a 480x854 portrait clip. Use adaptive to keep the source's own shape instead. One of: `16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `21:9`, `adaptive`. Default: "16:9".
duration	integer	no	Length of the generated clip in seconds, from 4 to 30. Leave at -1 (the default) to let the model choose a whole-second length close to the source footage. Billing counts this alongside the duration of the source clips you send. Default: -1.
generate_audio	boolean	no	Generate a synchronized soundtrack (dialogue, ambience and sound effects) together with the picture. Set false for a silent clip; the price is the same either way. Default: true.
reference_audios	array	no	Experimental: up to 10 short audio references for the generated soundtrack, addressed as @Audio1, @Audio2 and so on. Each clip 2-30 seconds, 30 seconds combined at most; WAV or MP3, under 15 MB each. The clips are accepted and add nothing to the price, but their effect on the finished soundtrack has not been verified - treat the field as experimental. They cannot be sent on their own: at least one reference video is always required.

Machine-readable: OpenAPI schema · llms.txt

Use Seedance 2.5 Video to Video from Claude & Cursor (MCP)

Point Claude Code, Claude Desktop, Cursor, or any MCP client at the ModelRunner MCP server and Seedance 2.5 Video to Video becomes a tool your assistant can call directly — it authorizes via OAuth (no API key in config) and runs this model with the run_model tool using the endpoint bytedance/seedance-v2.5/video-to-video.

MCP client config (Claude Desktop, Cursor)

{
  "mcpServers": {
    "modelrunner": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.modelrunner.run/mcp"]
    }
  }
}

Claude Code

claude mcp add --transport http modelrunner https://mcp.modelrunner.run/mcp

Then ask your assistant, for example: “Run bytedance/seedance-v2.5/video-to-video on ModelRunner to generate video”. MCP setup guide.

Model Detayları

Seedance 2.5 Video to Video, elinizdeki görüntüden yeni bir klip üretir — bir çekimi başka bir yerde yeniden sahneleyin, görünümünü değiştirin, kamera hareketini ve temposunu başka bir sahneye taşıyın ya da aksiyonu kaynağın bittiği yerin ötesine sürdürün. `reference_videos` alanına 1 ila 10 kaynak klip gönderin, isteğe bağlı olarak referans görsel ve ses ekleyin, sonucu tarif edin. Mevcut görüntüyü kabul eden tek Seedance 2.5 varyantı budur; fiyatı yalnızca geri aldığınıza göre değil, gönderdiğinize göre de ölçeklenen tek varyant da odur — bu yüzden her kaynağı işinize yarayan bölüme kırpın. Çıktı 480p ya da 720p; 24 fps’de 30 saniyeye kadar uzar ve ek ücret olmadan senkronize ses taşır. Kare biçimi size kalmış: 16:9 bir kaynak 9:16 dikey kesim olarak dönebilir, `adaptive` ise kaynağın kendi biçimini korur.

**Her referansı prompt’ta adıyla anın.** Referanslara konumlarına göre `@Video1`…`@Video10`, `@Image1`…`@Image30` ve `@Audio1`…`@Audio10` diye seslenilir; numaralar `reference_videos`, `reference_images` ve `reference_audios` içindeki sıralarından gelir. Prompt’un hiç anmadığı bir referans genellikle yok sayılır:

``` reference_videos: [ lantern-alley-night.mp4 ] reference_images: [ cherry-red-motorcycle.jpg ] prompt: "Match the camera movement and pacing of @Video1, but restage it on a sunlit desert roadside at midday; the rider pulls up on the motorcycle from @Image1." ```

## En uygun olduğu işler - Mevcut görüntüyü yeniden tarzlamak ya da yeniden sahnelemek — yeni mekân, mevsim ya da görünüm; kaynağın kamera hareketi ve temposu korunarak - Bir çekimi, kaynak görüntünün bittiği yerin ötesine uzatmak ya da sürdürmek - Dönüştürürken yeniden çerçevelemek — yatay bir kaynağı dikey kesime çevirmek ya da özgün biçimi korumak için `adaptive` - Bir kaynak klibi referans görseller ve seslerle tek üretimde harmanlamak; her birine adıyla seslenerek

## Şu durumlarda başka bir model seçin - Kaynak görüntünüz yoksa: `bytedance/seedance-v2.5/image-to-video` (tek durağan görsel), `bytedance/seedance-v2.5/first-last-frame` (bir keyframe çifti) ya da `bytedance/seedance-v2.5/text-to-video` (yalnızca prompt) kullanın - Referanslarınız var ama videonuz yoksa — `bytedance/seedance-v2.5/reference-to-video` daha ucuza gelir, yalnızca biten klibi fiyatlandırır - 720p’den fazlası gerekiyorsa — `bytedance/seedance-v2/video-to-video` 1080p’ye çıkar - Özgün dosyanın kare hassasiyetinde bir kurgusu gerekiyorsa — bu model yeni bir üretim döndürür; kareler, yüzler ve ekrandaki metin birebir korunmaz - Sabit bir seed ya da kilitli bir kamera gerekiyorsa — bu üretimde ikisi de yok

## İpuçları - Kaynak klipleri göndermeden önce kırpın: süreleri çıktıyla birlikte faturalanır, ayrıca uzun ya da kalabalık bir klip prompt’un denetimini zayıflatır - Sıra adrestir: `@Video1`, `reference_videos` içindeki ilk URL’dir - Hangi özelliği ödünç aldığınızı söyleyin — hareket, çerçeveleme, ışık — yalnızca "referanstaki gibi" demeyin - `reference_audios` deneyseldir: klipler kabul edilir ve ücretlendirilmez, ancak ses bandına etkisi doğrulanmamıştır

ModelRunner JavaScript client ile çalıştırmak için: ```js import { modelrunner } from "@modelrunner/client";

const result = await modelrunner.subscribe("bytedance/seedance-v2.5/video-to-video", { input: { prompt: "Match the camera movement and pacing of @Video1, but restage it on a sunlit desert roadside at midday; the rider pulls up on the cherry-red motorcycle from @Image1", reference_videos: ["https://media.modelrunner.ai/6m0I6xs8DL2oaJhQoD8x9.mp4"], reference_images: ["https://media.modelrunner.ai/6M3U2hs7xUtWJGZmjkvuN.jpeg"], resolution: "720p", aspect_ratio: "16:9", duration: 6, generate_audio: true, }, }); ```

Seedance 2.5 Video to Video API

Model girdisi

Input

Drag & drop videos here

Drag & drop images here

Additional Settings

Drag & drop audio files here

Model çıktısı

Output

Örnek istekler

Örnekler

Model fiyatlandırması

Fiyatlandırma

Seedance 2.5 Video to Video API

cURL

JavaScript

Python

Input parameters

Use Seedance 2.5 Video to Video from Claude & Cursor (MCP)

MCP client config (Claude Desktop, Cursor)

Claude Code

Model Detayları

Model Detayları