The TwelveLabs class provides methods to analyze videos synchronously and generate text based on their content. These methods return results immediately in the response.

When to use these methods:

Analyze videos up to 1 hour, or images
Retrieve immediate results without polling for task completion
Stream text fragments in real time for immediate processing and feedback

Do not use these methods for:

Videos longer than 1 hour. Use the analyze_async.tasks.create method instead.
Video segmentation with custom segment definitions. Use the analyze_async.tasks.create method instead.

Sync analysis

Description: This method analyzes a video or one or more images and returns the results directly in the response. Each request must contain a video or one or more images, but not both. You can use general analysis (prompt-based text generation) with either media type.

Input requirements

Videos

Minimum duration: 1 second
Maximum duration: 1 hour
Formats: FFmpeg supported formats
Resolution: 360x360 to 5184x2160 pixels
Aspect ratio: Between 1:1 and 1:2.4, or between 2.4:1 and 1:1.

Images

You can provide one to twenty images per request.
Formats: JPEG, PNG, WebP, GIF, and BMP.
Maximum size: 20 MB per image.
Maximum pixel count: 16,777,216 pixels per image (width × height).

On the Free plan in legacy billing, you have a total of 600 minutes (10 hours) shared across indexing, analysis, and segmentation. For details, see the Video hours and video count limits section. In the new billing, the allowances of your plan apply in place of this limit. See Plans, allowances, and overage.

Note

This method is rate-limited. For details, see the Rate limits page.

Function signature and example:

def analyze(
    self,
    *,
    model_name: typing.Optional[AnalyzeRequestModelName] = OMIT,
    video: typing.Optional[VideoContext] = OMIT,
    image: typing.Optional[typing.Sequence[AnalyzeImageInput]] = OMIT,
    prompt: typing.Optional[AnalyzeTextPrompt] = OMIT,
    prompt_v_2: typing.Optional[AnalyzePromptV2] = OMIT,
    temperature: typing.Optional[AnalyzeTemperature] = OMIT,
    response_format: typing.Optional[SyncResponseFormat] = OMIT,
    max_tokens: typing.Optional[int] = OMIT,
    start_time: typing.Optional[float] = OMIT,
    end_time: typing.Optional[float] = OMIT,
    request_options: typing.Optional[RequestOptions] = None,
) -> NonStreamAnalyzeResponse:

Parameters:

Name	Type	Required	Description
`model_name`	`typing.Optional[AnalyzeRequestModelName]`	No	The video understanding model to use for analysis. Values: - `"pegasus1.6"`: For details about this version, see the Pegasus 1.6 page. - `"pegasus1.5"`: For details about this version, see the Pegasus 1.5 page. Default: `"pegasus1.5"`
`video`	`typing.Optional[VideoContext]`	No	An object specifying the source of the video content. Include exactly one source. Mutually exclusive with the `image` parameter. See VideoContext.
`image`	`typing.Optional[typing.Sequence[AnalyzeImageInput]]`	No	A list of up to twenty objects containing the images to analyze. For each image, include exactly one source. Requires Pegasus 1.6. Using any other model returns a `parameter_invalid` error. Mutually exclusive with the `video` and `prompt_v_2` parameters. The `prompt` parameter is required when you provide images. See AnalyzeImageInput for details.
`prompt`	`typing.Optional[str]`	No	A text prompt that guides the model on the desired format or content. To include reference images in your prompt, use the `prompt_v_2` parameter instead. Mutually exclusive with the `prompt_v_2` parameter.
`prompt_v_2`	`typing.Optional[AnalyzePromptV2]`	No	A structured prompt that uses the `<@name>` placeholders to reference images. Mutually exclusive with the `prompt` and `image` parameters. See AnalyzePromptV2.
`temperature`	`typing.Optional[float]`	No	Controls the randomness of the text output. Default: 0.2, Min: 0, Max: 1
`start_time`	`typing.Optional[float]`	No	Start of the analysis window, as an absolute timestamp in seconds, based on the internal metadata of the video. Use with `end_time` to analyze only a portion of the video. If omitted, defaults to the internal start time of the video. Most videos start at 0, but some (for example, from cameras or broadcast recordings) may have a non-zero start time. To find the value, run `ffprobe -v error -show_entries format=start_time,duration -of default=noprint_wrappers=1 your_video.mp4`. Must be less than `end_time` and the video duration. The window (`end_time - start_time`) must be at least 1 second.
`end_time`	`typing.Optional[float]`	No	End of the analysis window, as an absolute timestamp in seconds, based on the internal metadata of the video. Use with `start_time` to analyze only a portion of the video. If omitted, defaults to the internal start time of the video plus its duration. Most videos start at 0, but some (for example, from cameras or broadcast recordings) may have a non-zero start time. To find the value, run `ffprobe -v error -show_entries format=start_time,duration -of default=noprint_wrappers=1 your_video.mp4`. Must be greater than `start_time` and less than or equal to the video duration. The window (`end_time - start_time`) must be at least 1 second.
`request_options`	`typing.Optional[RequestOptions]`	No	Request-specific configuration.

The SyncResponseFormat class contains the following properties:

Name	Type	Description
`type`	`SyncResponseFormatType`	The response format to use. Value: `"json_schema"` (structured JSON conforming to a provided schema).
`json_schema`	`Dict[str, Optional[Any]]`	Contains the JSON schema that defines the response structure. The schema must adhere to the JSON Schema Draft 2020-12 specification. For details, see the `json_schema` parameter in the API Reference section.

VideoContext

The VideoContext type specifies the source of the video content. Provide exactly one of the following:

Class	Field	Type	Description
`VideoContext_Url`	`url`	`str`	The publicly accessible URL of the video file or HLS manifest. Use direct links to raw media files, or the URL of a VOD HLS manifest. Live video streams are rejected with a `400` error. Video hosting platforms and cloud storage sharing links are not supported. For HLS sources, if the duration cannot be determined, the platform calculates it from the manifest and the duration limits apply to that value.
`VideoContext_AssetId`	`asset_id`	`str`	The unique identifier of an asset from a direct or multipart upload. The asset status must be `ready`. Use `assets.retrieve` to check the status.
`VideoContext_Base64String`	`base_64_string`	`str`	The base64-encoded video data. The maximum size is 30MB.

AnalyzeImageInput

The AnalyzeImageInput class defines an image to analyze. Provide exactly one of the url, asset_id, or base64_string fields.

Name	Type	Required	Description
`name`	`str`	Yes	The name of this image. Must be unique within the `image` array. Can contain only letters, digits, and underscores. To refer to a specific image in your prompt, use its name as the `<@name>` placeholder. Reference every image or none of them: - If the prompt references every image, the platform analyzes each one (Example: `Compare <@before> with <@after> and describe what changed.`). - If the prompt references none of them, the platform analyzes every image you provided (Example: `List every visible feature in these images.`). - If the prompt references only some of the images, the platform returns a `400` error.
`url`	`Optional[str]`	No	A publicly accessible HTTPS URL of the image. The platform rejects HTTP URLs. When you send the request, the platform verifies that the URL is reachable and that the image is within the size, pixel count, and format limits. If either check fails, the platform returns a `400` error.
`asset_id`	`Optional[str]`	No	The unique identifier of an uploaded asset.
`base_64_string`	`Optional[str]`	No	Base64-encoded image data. The maximum decoded size is 20 MB. Supported image types: `image/jpeg`, `image/png`, `image/webp`, `image/gif`, and `image/bmp`. Standard and unpadded base64 encodings are both accepted.

AnalyzePromptV2

The AnalyzePromptV2 class defines a structured prompt that uses <@name> placeholders to reference images. Mutually exclusive with the prompt and image parameters.

Name	Type	Required	Description
`input_text`	`str`	Yes	The text of the prompt. Use `<@name>` placeholders to reference images declared in `media_sources` (Example: `"Is there a <@tiger-1> in the video?"`). This text counts toward the context window.
`media_sources`	`Optional[List[SmeMediaSource]]`	No	Reference images for the `<@name>` placeholders in the prompt. Maximum 4 sources. See SmeMediaSource.

SmeMediaSource

A reference image that provides visual context. Provide exactly one of the url, asset_id, or base_64_string fields.

Name	Type	Required	Description
`name`	`str`	Yes	A descriptive name for this media source.
`media_type`	`SmeMediaSourceMediaType`	Yes	The media type. Value: `"image"`.
`url`	`Optional[str]`	No	A publicly accessible HTTPS URL of the image.
`asset_id`	`Optional[str]`	No	The unique identifier of an uploaded asset.
`base_64_string`	`Optional[str]`	No	Base64-encoded image data. The maximum decoded size is 32 MB.

Return value: Returns a NonStreamAnalyzeResponse object containing the generated text.

The NonStreamAnalyzeResponse class contains the following properties:

Name	Type	Description
`id`	`Optional[str]`	Unique identifier of the response.
`data`	`Optional[str]`	The generated text based on the prompt you provided.
`finish_reason`	`typing.Optional[FinishReason]`	The reason the generation stopped. Values: - `null`: The generation has not finished yet. - `"stop"`: The generation reached the end of the output text. - `"length"`: The generation reached the maximum response length or the context window. When you set the `response_format` parameter, the output may be truncated and fail to parse. The partial output is in `data`, and a warning is in the `error` field.
`usage`	`Optional[TokenUsage]`	The number of tokens used in the generation.

The TokenUsage class contains the following properties:

Name	Type	Description
`output_tokens`	`int`	The number of tokens in the generated text.
`input_tokens`	`Optional[int]`	The number of tokens the input consumed. Together with `output_tokens`, this value must fit within the context window.

API Reference: Sync analysis.

Related guide: Analyze videos and images.

Sync analysis with streaming responses

Input requirements

Videos

Minimum duration: 1 second
Maximum duration: 1 hour
Formats: FFmpeg supported formats
Resolution: 360x360 to 5184x2160 pixels
Aspect ratio: Between 1:1 and 1:2.4, or between 2.4:1 and 1:1.

Images

You can provide one to twenty images per request.
Formats: JPEG, PNG, WebP, GIF, and BMP.
Maximum size: 20 MB per image.
Maximum pixel count: 16,777,216 pixels per image (width × height).

Note

This method is rate-limited. For details, see the Rate limits page.

Function signature and example:

def analyze_stream(
    self,
    *,
    model_name: typing.Optional[AnalyzeStreamRequestModelName] = OMIT,
    video: typing.Optional[VideoContext] = OMIT,
    image: typing.Optional[typing.Sequence[AnalyzeImageInput]] = OMIT,
    prompt: typing.Optional[AnalyzeTextPrompt] = OMIT,
    prompt_v_2: typing.Optional[AnalyzePromptV2] = OMIT,
    temperature: typing.Optional[AnalyzeTemperature] = OMIT,
    response_format: typing.Optional[SyncResponseFormat] = OMIT,
    max_tokens: typing.Optional[int] = OMIT,
    start_time: typing.Optional[float] = OMIT,
    end_time: typing.Optional[float] = OMIT,
    request_options: typing.Optional[RequestOptions] = None,
) -> typing.Iterator[StreamAnalyzeResponse]

Parameters:

Name	Type	Required	Description
`model_name`	`typing.Optional[AnalyzeRequestModelName]`	No	The video understanding model to use for analysis. Values: - `"pegasus1.6"`: For details about this version, see the Pegasus 1.6 page. - `"pegasus1.5"`: For details about this version, see the Pegasus 1.5 page. Default: `"pegasus1.5"`
`video`	`typing.Optional[VideoContext]`	No	An object specifying the source of the video content. Include exactly one source. Mutually exclusive with the `image` parameter. See VideoContext.
`image`	`typing.Optional[typing.Sequence[AnalyzeImageInput]]`	No	A list of up to twenty objects containing the images to analyze. For each image, include exactly one source. Requires Pegasus 1.6. Using any other model returns a `parameter_invalid` error. Mutually exclusive with the `video` and `prompt_v_2` parameters. The `prompt` parameter is required when you provide images. See AnalyzeImageInput for details.
`prompt`	`typing.Optional[str]`	No	A text prompt that guides the model on the desired format or content. To include reference images in your prompt, use the `prompt_v_2` parameter instead. Mutually exclusive with the `prompt_v_2` parameter.
`prompt_v_2`	`typing.Optional[AnalyzePromptV2]`	No	A structured prompt that uses the `<@name>` placeholders to reference images. Mutually exclusive with the `prompt` and `image` parameters. See AnalyzePromptV2.
`temperature`	`typing.Optional[float]`	No	Controls the randomness of the text output. Default: 0.2, Min: 0, Max: 1
`start_time`	`typing.Optional[float]`	No	Start of the analysis window, as an absolute timestamp in seconds, based on the internal metadata of the video. Use with `end_time` to analyze only a portion of the video. If omitted, defaults to the internal start time of the video. Most videos start at 0, but some (for example, from cameras or broadcast recordings) may have a non-zero start time. To find the value, run `ffprobe -v error -show_entries format=start_time,duration -of default=noprint_wrappers=1 your_video.mp4`. Must be less than `end_time` and the video duration. The window (`end_time - start_time`) must be at least 1 second.
`end_time`	`typing.Optional[float]`	No	End of the analysis window, as an absolute timestamp in seconds, based on the internal metadata of the video. Use with `start_time` to analyze only a portion of the video. If omitted, defaults to the internal start time of the video plus its duration. Most videos start at 0, but some (for example, from cameras or broadcast recordings) may have a non-zero start time. To find the value, run `ffprobe -v error -show_entries format=start_time,duration -of default=noprint_wrappers=1 your_video.mp4`. Must be greater than `start_time` and less than or equal to the video duration. The window (`end_time - start_time`) must be at least 1 second.
`request_options`	`typing.Optional[RequestOptions]`	No	Request-specific configuration.

For details about VideoContext, AnalyzeImageInput, AnalyzePromptV2, and SmeMediaSource, see the Sync analysis section above.

The SyncResponseFormat class contains the following properties:

Name	Type	Description
`type`	`SyncResponseFormatType`	The response format to use. Value: `"json_schema"` (structured JSON conforming to a provided schema).
`json_schema`	`Dict[str, Optional[Any]]`	Contains the JSON schema that defines the response structure. The schema must adhere to the JSON Schema Draft 2020-12 specification. For details, see the `json_schema` parameter in the API Reference section.

Return value: Returns an iterator of StreamAnalyzeResponse objects. Each response can be a StreamAnalyzeResponse_StreamStart, StreamAnalyzeResponse_TextGeneration, or StreamAnalyzeResponse_StreamEnd.

The StreamAnalyzeResponse_StreamStart class contains the following properties:

Name	Type	Description
`event_type`	`typing.Literal["stream_start"]`	This field is always set to `stream_start` for this event.
`metadata`	`Optional[StreamStartResponseMetadata]`	An object containing metadata about the stream.

The StreamAnalyzeResponse_TextGeneration class contains the following properties:

Name	Type	Description
`event_type`	`typing.Literal["text_generation"]`	This field is always set to `text_generation` for this event.
`text`	`Optional[str]`	A fragment of the generated text. Text fragments may be split at arbitrary points, not necessarily at word or sentence boundaries.

The StreamAnalyzeResponse_StreamEnd class contains the following properties:

Name	Type	Description
`event_type`	`typing.Literal["stream_end"]`	This field is always set to `stream_end` for this event.
`finish_reason`	`typing.Optional[FinishReason]`	The reason the generation stopped. Values: - `null`: The generation has not finished yet. - `"stop"`: The generation reached the end of the output text. - `"length"`: The generation reached the maximum response length or the context window. When you set the `response_format` parameter, the output may be truncated and fail to parse. The partial output is in `data`, and a warning is in the `error` field.
`metadata`	`typing.Optional[StreamEndResponseMetadata]`	An object containing metadata about the stream.

The StreamStartResponseMetadata class contains the following properties:

Name	Type	Description
`generation_id`	`Optional[str]`	A unique identifier for the generation session.

The StreamEndResponseMetadata class contains the following properties:

Name	Type	Description
`generation_id`	`Optional[str]`	The same unique identifier provided in the `stream_start` event.
`usage`	`Optional[TokenUsage]`	The number of tokens used in the generation.

The TokenUsage class contains the following properties:

Name	Type	Description
`output_tokens`	`int`	The number of tokens in the generated text.
`input_tokens`	`Optional[int]`	The number of tokens the input consumed. Together with `output_tokens`, this value must fit within the context window.

API Reference: Sync analysis.

Related guide: Analyze videos and images.

Error codes

This section lists the most common error messages you may encounter while analyzing videos.

token_limit_exceeded
- The request exceeds the context window and cannot be processed. Reduce the prompt length, use a shorter video, or lower the max_tokens value.
Live video streams are not supported. Please provide a URL that is not a live stream.
- The URL points to a live HLS stream. Only VOD manifests are supported. Provide the URL of a VOD manifest.
output truncated: the generation reached the configured max_tokens. The partial output is returned; raise max_tokens if you need a longer response.
- The response reached the maximum response length in general mode. The platform returns the partial output and sets finish_reason to "length". This message appears in the error.message field.
output truncated: combined input and output tokens reached the model's context limit. The partial output is returned; consider reducing input size (shorter prompt, smaller video clip, fewer media bindings) or lowering max_tokens.
- The request reached the context window in general mode. The platform returns the partial output and sets finish_reason to "length". This message appears in the error.message field.
analysis failed: the time_based_metadata output reached the configured max_tokens before a complete result was produced. Raise max_tokens if it is below the per-model maximum; otherwise narrow the request (fewer segment_definitions or fields, a larger min_segment_duration, or a shorter analysis window).
- The response reached the maximum response length in time_based_metadata mode. The task fails with status set to "failed", and no partial output is returned. This message appears in the error.message field.
analysis failed: the time_based_metadata output reached the model's context limit (combined input and output tokens) before a complete result was produced. Narrow the request (fewer segment_definitions or fields, a larger min_segment_duration, a shorter analysis window, fewer media bindings) or lower max_tokens to leave more room for the input.
- The request reached the context window in time_based_metadata mode. The task fails with status set to "failed", and no partial output is returned. This message appears in the error.message field.
user_canceled
- An async analysis task was canceled through the task cancellation endpoint. The task has status set to "canceled", and no result is returned.
batch_canceled
- A queued analysis task was canceled because its batch was canceled. The task has status set to "canceled", and no result is returned.
batch_expired
- A queued analysis task was canceled because its batch expired. The task has status set to "canceled", and no result is returned.
batch_processing_timeout
- A processing analysis task failed after reaching its batch’s processing deadline. The task has status set to "failed", and no result is returned.
index_not_supported_for_generate
- You can only summarize videos uploaded to an index with an engine from the Pegasus family enabled.

For a list of general errors that apply to all endpoints, see the Error codes page.