Seedance 2.5 creates video using a written scene and multimodal references. The exact selector doubao-seedance-2-5-260628 has its own documented operation and example. That operation describes video generation, video editing and video extension, with images, video, audio and text as input media. Use the references to communicate subject appearance, action, camera movement, style and sound, then explain the result to create from them.
The input combinations include text alone, images or video with optional text, images with audio, video with audio, and images, video and audio together. The listed combinations attach audio to visual media rather than describing an audio-only request. Organize each item by its role: an image establishes a subject or look, a video demonstrates movement or an existing scene, and audio supplies a sound reference. State which role matters for the requested shot so a background reference is not mistaken for the principal character.
The exact 2.5 example is a product-advertising sequence. It supplies an image and several video references, then describes an opening focus, ordered product arrangements, detail shots, a slow-motion action and a closing brand composition. The request assigns reference_image and reference_video roles and uses individual references for composition, movement, camera work and impact. This gives a practical structure for a campaign concept: plan the visible beats, identify the reference for each beat and specify the final arrangement.
A written sequence can distinguish what happens from how it is filmed. Describe the product action first, then the shot size, camera route, pacing and transition to the next view. If one reference defines the product and another demonstrates a camera movement, name those separate responsibilities. For a scene with several objects, explain their positions and the order of interactions. Inspect the generated result for the actual product shape, arrangement and movement rather than assuming every reference detail appears unchanged.
For an edit, identify the existing video and the element to revise, such as the setting, an action or the visual style, while stating the continuity to retain. For an extension, describe how the scene should continue from the supplied footage. Review the join as well as the new segment: subject appearance, direction of movement, camera position and sound timing all affect whether the result reads as one scene. Keep the original video and completed revision together so the change can be checked against the intended continuity.
The documented 2.5 multimodal example sets generate_audio=true and requests a fifteen-second sequence with 16:9 framing. Those values describe that example rather than a universal duration ceiling or reference-asset limit. Give the requested scene a clear sound intent, including the relationship between music, effects and visible action, then evaluate the actual audiovisual output. A sound reference and the generated soundtrack should be reviewed together for the intended pacing and performance.
Choose duration, framing and resolution from the settings offered for this exact profile. Keep the requested configuration with the prompt and compare it with the actual returned task fields. Use the returned duration, resolution and ratio when preparing the completed clip for its destination. Preserve the exact profile identifier in the revision record so a later comparison uses the same selected edition.
Video-token usage is calculated as (input-video duration + output-video duration) × output width × output height × output frame rate / 1024. Requests containing input video and those without input video have distinct billing categories. The task query reports usage.completion_tokens and usage.total_tokens; retain these reported quantities with the completed clip.
Generation is asynchronous. Submission returns an identifier, and the configured task flow provides status followed by content.video_url after success. Save the completed clip together with the source media and the sequence brief. Review all planned beats, object interactions, character appearance, framing and sound alignment across the whole result. For a revision, identify the concrete missed detail and preserve the successful parts of the scene rather than replacing the entire brief with unrelated instructions.
", "aboutText": "Seedance 2.5 creates video using a written scene and multimodal references. The exact selector doubao-seedance-2-5-260628 has its own documented operation and example. That operation describes video generation, video editing and video extension, with images, video, audio and text as input media. Use the references to communicate subject appearance, action, camera movement, style and sound, then explain the result to create from them.\n\nThe input combinations include text alone, images or video with optional text, images with audio, video with audio, and images, video and audio together. The listed combinations attach audio to visual media rather than describing an audio-only request. Organize each item by its role: an image establishes a subject or look, a video demonstrates movement or an existing scene, and audio supplies a sound reference. State which role matters for the requested shot so a background reference is not mistaken for the principal character.\n\nThe exact 2.5 example is a product-advertising sequence. It supplies an image and several video references, then describes an opening focus, ordered product arrangements, detail shots, a slow-motion action and a closing brand composition. The request assigns reference_image and reference_video roles and uses individual references for composition, movement, camera work and impact. This gives a practical structure for a campaign concept: plan the visible beats, identify the reference for each beat and specify the final arrangement.\n\nA written sequence can distinguish what happens from how it is filmed. Describe the product action first, then the shot size, camera route, pacing and transition to the next view. If one reference defines the product and another demonstrates a camera movement, name those separate responsibilities. For a scene with several objects, explain their positions and the order of interactions. Inspect the generated result for the actual product shape, arrangement and movement rather than assuming every reference detail appears unchanged.\n\nFor an edit, identify the existing video and the element to revise, such as the setting, an action or the visual style, while stating the continuity to retain. For an extension, describe how the scene should continue from the supplied footage. Review the join as well as the new segment: subject appearance, direction of movement, camera position and sound timing all affect whether the result reads as one scene. Keep the original video and completed revision together so the change can be checked against the intended continuity.\n\nThe documented 2.5 multimodal example sets generate_audio=true and requests a fifteen-second sequence with 16:9 framing. Those values describe that example rather than a universal duration ceiling or reference-asset limit. Give the requested scene a clear sound intent, including the relationship between music, effects and visible action, then evaluate the actual audiovisual output. A sound reference and the generated soundtrack should be reviewed together for the intended pacing and performance.\n\nChoose duration, framing and resolution from the settings offered for this exact profile. Keep the requested configuration with the prompt and compare it with the actual returned task fields. Use the returned duration, resolution and ratio when preparing the completed clip for its destination. Preserve the exact profile identifier in the revision record so a later comparison uses the same selected edition.\n\nVideo-token usage is calculated as (input-video duration + output-video duration) × output width × output height × output frame rate / 1024. Requests containing input video and those without input video have distinct billing categories. The task query reports usage.completion_tokens and usage.total_tokens; retain these reported quantities with the completed clip.\n\nGeneration is asynchronous. Submission returns an identifier, and the configured task flow provides status followed by content.video_url after success. Save the completed clip together with the source media and the sequence brief. Review all planned beats, object interactions, character appearance, framing and sound alignment across the whole result. For a revision, identify the concrete missed detail and preserve the successful parts of the scene rather than replacing the entire brief with unrelated instructions.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "1278dd07e7e2137847134d174ec799a9b93b933276abda44afc0b79e1a23f690", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "doubao-seedance-2-5-260628Seedance 2.5 creates video from text and image, video or audio references, with documented editing and extension workflows. Its exact 2.5 example combines a product image with several motion references and enables generated sound for a coordinated advertising sequence.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Seedance 2.5 creates video using a written scene and multimodal references. The exact selector doubao-seedance-2-5-260628 has its own documented operation and example. That operation describes video generation, video editing and video extension, with images, video, audio and text as input media. Use the references to communicate subject appearance, action, camera movement, style and sound, then explain the result to create from them.
The input combinations include text alone, images or video with optional text, images with audio, video with audio, and images, video and audio together. The listed combinations attach audio to visual media rather than describing an audio-only request. Organize each item by its role: an image establishes a subject or look, a video demonstrates movement or an existing scene, and audio supplies a sound reference. State which role matters for the requested shot so a background reference is not mistaken for the principal character.
The exact 2.5 example is a product-advertising sequence. It supplies an image and several video references, then describes an opening focus, ordered product arrangements, detail shots, a slow-motion action and a closing brand composition. The request assigns reference_image and reference_video roles and uses individual references for composition, movement, camera work and impact. This gives a practical structure for a campaign concept: plan the visible beats, identify the reference for each beat and specify the final arrangement.
A written sequence can distinguish what happens from how it is filmed. Describe the product action first, then the shot size, camera route, pacing and transition to the next view. If one reference defines the product and another demonstrates a camera movement, name those separate responsibilities. For a scene with several objects, explain their positions and the order of interactions. Inspect the generated result for the actual product shape, arrangement and movement rather than assuming every reference detail appears unchanged.
For an edit, identify the existing video and the element to revise, such as the setting, an action or the visual style, while stating the continuity to retain. For an extension, describe how the scene should continue from the supplied footage. Review the join as well as the new segment: subject appearance, direction of movement, camera position and sound timing all affect whether the result reads as one scene. Keep the original video and completed revision together so the change can be checked against the intended continuity.
The documented 2.5 multimodal example sets generate_audio=true and requests a fifteen-second sequence with 16:9 framing. Those values describe that example rather than a universal duration ceiling or reference-asset limit. Give the requested scene a clear sound intent, including the relationship between music, effects and visible action, then evaluate the actual audiovisual output. A sound reference and the generated soundtrack should be reviewed together for the intended pacing and performance.
Choose duration, framing and resolution from the settings offered for this exact profile. Keep the requested configuration with the prompt and compare it with the actual returned task fields. Use the returned duration, resolution and ratio when preparing the completed clip for its destination. Preserve the exact profile identifier in the revision record so a later comparison uses the same selected edition.
Video-token usage is calculated as (input-video duration + output-video duration) × output width × output height × output frame rate / 1024. Requests containing input video and those without input video have distinct billing categories. The task query reports usage.completion_tokens and usage.total_tokens; retain these reported quantities with the completed clip.
Generation is asynchronous. Submission returns an identifier, and the configured task flow provides status followed by content.video_url after success. Save the completed clip together with the source media and the sequence brief. Review all planned beats, object interactions, character appearance, framing and sound alignment across the whole result. For a revision, identify the concrete missed detail and preserve the successful parts of the scene rather than replacing the entire brief with unrelated instructions.
POST /v1/videostext, image, video, audiovideo| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits doubao-seedance-2-5-260628 | Default— |
Fieldcontent | Typearray | RequirementRequired | Values & limits Conditions: Native text/media objects; outer array must be non-empty. Nested node types are not validated by this Ailoot generic constructor. | Default— |
Fieldgenerate_audio | Typeboolean | RequirementOptional | Values & limits Conditions: 2.5 example sets true but copied selector prose lists only 2.0/fast/1.5; no wider setting/default guarantee. | Default— |
Fieldduration | Typeinteger | RequirementOptional | Values & limits Conditions: Literal2.5 example 15 is not a maximum or default. | Default— |
Fieldresolution | Typestring | RequirementOptional | Values & limits Conditions: No 2.5-specific default/enum inferred from copied descriptions. | Default— |
Fieldratio | Typestring | RequirementOptional | Values & limits Conditions: Adaptive explicitly scoped to 2.0/fast/1.5; 2.5 example 16:9 only. | Default— |
Fieldwatermark | Typeboolean | RequirementOptional | Values & limits | Defaultfalse |
Fieldseed | Typeinteger | RequirementOptional | Values & limits Minimum: -1 Maximum: 4,294,967,295 | Default-1 |
Fieldexecution_expires_after | Typeinteger | RequirementOptional | Values & limits Minimum: 3,600 Maximum: 259,200 Unit: seconds | Default172,800 |
Fieldservice_tier | Typestring | RequirementOptional | Values & limits Conditions: 2.0/fast exclude offline flex; 2.5 scope is not established. | Default— |
Fieldsafety_identifier | Typestring | RequirementOptional | Values & limits | Default— |
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractcontent, model |
Parameteroptional | Contractgenerate_audio, duration, resolution, ratio, watermark, seed, execution_expires_after, service_tier, safety_identifier |
Parametermaterial_types | Contracttext, image, video, audio |
Parameteradditional_properties | Contracttrue |
Parametervalidation_scope | ContractAiloot checks the required field types and a non-empty model/content array. Other request fields are preserved for the selected model. |
Parametermaterial_combinations | Contracttext, image, video, image, audio, image, video, video, audio, image, video, audio |
Parameteroptional_text_with_media | Contracttrue |
Parameterreference_media_count_and_size_limits | Contractnull |
Method aliases: /v1/videos/generations
Lifecycle: submit_once_poll_original
GET /v1/videos/{request_id}
GET /v1/videos/generations/{request_id}
GET /v1/videos/{request_id}/content
GET /v1/videos/generations/{request_id}/content
Create once and retain returned public id/polling_url.
Poll the original job; download actual owned content.video_url after succeeded.
The displayed rates estimate the cost before generation. The final charge is calculated from the cost recorded for the original generation. Measured output duration and forecast quantities are not the final billable quantity.
Tariff references do not add request options. Supported parameters are listed for the selected method.
Match each charge to the conditions shown. Use one rate for each set of alternative configurations.
| Configuration | Rate schedule | Forecast quantity | Billing unit | Charge |
|---|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Forecast quantity— | Billing unit— | Charge Input · Input tokens · Estimate for comparison only — / 1K tokens Output · Output tokens · Estimate for comparison only — / 1K tokens Input · Cache read · Estimate for comparison only — / 1K tokens Input · Cache write · Estimate for comparison only — / 1K tokens Input · Cache write, 1 hour · Estimate for comparison only — / 1K tokens |
| ConfigurationInput contains video. | Rate schedule— | Forecast quantity— | Billing unitper 1,000 video tokens | Charge Output · Estimate for comparison only $0.012 per 1,000 video tokens |
| ConfigurationInput contains no video. | Rate schedule— | Forecast quantity— | Billing unitper 1,000 video tokens | Charge Output · Estimate for comparison only $0.02 per 1,000 video tokens |
| ConfigurationEstimated generation cost | Rate schedule— | Forecast quantityone generation estimate | Billing unit— | Charge Output · Estimated generation cost · Estimate for comparison only $0.822402 estimate before generation This estimate is not an additional charge or a fixed rate for the completed generation. |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\ncat > request.json <<'JSON_REQUEST'\n{"model":"doubao-seedance-2-5-260628","content":[{"type":"text","text":"女孩抱着狐狸，女孩睁开眼，温柔地看向镜头，狐狸友善地抱着，镜头缓缓拉出，女孩的头发被风吹动"}]}\nJSON_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/videos \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\nset -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\npython3 - <<'PY_POLL'\nimport json\nfrom pathlib import Path\nfrom urllib.parse import quote,urljoin,urlsplit\nr=json.loads(Path("result.json").read_text())\ni=r.get("id")\nif not isinstance(i,str) or not i.startswith("native302_"):raise SystemExit("Keep original response/headers/id; do not resubmit creation.")\nu=urljoin("https://api.ailoot.ai",r.get("polling_url") or "/v1/videos/"+quote(i,safe=""))\np=urlsplit(u)\nif p.scheme!="https" or p.netloc!="api.ailoot.ai" or not p.path.startswith("/v1/videos/"):raise SystemExit("Use actual Ailoot lookup path.")\nPath("poll-url.txt").write_text(u)\nPY_POLL\ncurl --fail-with-body --silent --show-error "$(cat poll-url.txt)" -H "Authorization: Bearer $AILOOT_API_KEY" -D status-headers.txt -o status.json\nset -eu\npython3 - <<'PY_ASSETS'\nimport json\nfrom pathlib import Path\nfrom urllib.parse import urlsplit\nr=json.loads(Path("status.json").read_text())\nif r.get("status")!="succeeded":raise SystemExit("Poll same original job until an owned result is returned.")\nu=r.get("content",{}).get("video_url")\nurls=[u] if isinstance(u,str) else u\nif not isinstance(urls,list) or not urls:raise SystemExit("No completed video URL.")\nfor u in urls:\n if not isinstance(u,str):raise SystemExit("Invalid returned URL.")\n p=urlsplit(u)\n if p.scheme!="https" or p.netloc!="api.ailoot.ai" or not p.path.startswith("/v1/media/assets/") or not p.path.endswith("/content") or not p.query:raise SystemExit("Wait for actual Ailoot-owned signed result.")\nPath("asset-urls.txt").write_text("\\n".join(urls)+"\\n")\nPY_ASSETS\nindex=0\nwhile IFS= read -r asset_url; do\n curl --fail-with-body --silent --show-error "$asset_url" -o "video-${index}.bin"\n index=$((index + 1))\ndone < asset-urls.txt\nNative JSON with public id/polling_url. After publication content.video_url contains Ailoot-owned signed URL(s); before publication selected output URLs are blanked.
GET /v1/videos/{request_id}GET /v1/videos/generations/{request_id}GET /v1/videos/{request_id}/contentGET /v1/videos/generations/{request_id}/contentSave every completed video or returned file; retrieve every actual returned asset URL.
Retain every actual URL in content.video_url; index must identify a real returned output.
Preserve original response/headers/id and poll that operation; never resubmit unknown creation.
| Field | Type |
|---|---|
Fieldid | Typestring |
Fieldprovider_task_id | Typestring |
Fieldpolling_url | Typestring |
Fieldstatus | Typestring when supplied |
Fieldcontent.video_url | Typestring|string[] |
Fieldusage | Typeobject when supplied |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid JSON/model/required field types or content index. | Next step |
Response401 · authentication_error | ConditionInvalid Ailoot API key. | Next step |
Response404 · not_found_error | ConditionTask/output absent or not owned by same user/API key/group. | Next step |
Response502 · result_pending_check | ConditionOriginal submit/status incomplete; preserve original ID, no resubmit. | Next step |
Response503 · result_pending_check | ConditionOriginal media/route/settlement is being checked; continue original operation. | Next step |
Ailoot Ultra doubao-seedance-2-5-260628
/v1/videos/generations