A “poster frame” usually means one still image shown before playback; seek-preview thumbnails are a repeated set of images viewers move through while scrubbing. AWS MediaConvert can make either a sequence of JPEG captures or image assets associated with supported streaming outputs, but its documentation does not describe a separate one-off poster-frame feature.
Use FRAME_CAPTURE when you need JPEG files to choose from or process elsewhere. Use image-based trick-play generation when a compatible streaming output and player need preview images while seeking. The right path depends on how the image will be used, not just what you call it.
Choose between a poster still and seek previews
A poster still is a single representative image. It might appear in a video library, on a channel page, or before a player starts. MediaConvert’s standalone capture output creates numbered JPEGs; you can select one as a poster candidate in a later step. The capture feature does not itself promise to choose the best frame or deliver exactly one file at a particular arbitrary timestamp.
Trick-play images serve a different job. They provide visual previews along the playback timeline, helping a viewer locate a scene while seeking. Depending on the streaming output type and selected mode, MediaConvert can create tiled thumbnails, full-frame images, or both, and can associate those assets with manifests where supported.
| Need | Standalone FRAME_CAPTURE |
Image-based trick play |
|---|---|---|
| One still to use as a poster candidate | Capture JPEGs, then select one | Not a substitute for a dedicated chosen poster |
| Preview images while seeking | Not its purpose | Intended for supported streaming outputs and players |
| Image output | Numbered JPEG capture files | Tiled thumbnails and/or full-frame images, depending on settings |
| Timing control | Rational capture cadence and maximum capture count | IDR-following, custom interval, or segment-following cadence where available |
| Main verification | Match the numbered file to the intended moment | Check manifests, image layout, and target-player support |
If you are preparing a recorded programme that will later be broadcast continuously, a chosen still and an always-updating seek preview still solve different problems. For the broader broadcast workflow, see this guide to livestreaming pre-recorded videos on YouTube around the clock. MediaConvert makes media assets; it does not by itself decide how your YouTube broadcast is presented or kept running.
Prepare the MediaConvert job and output
Start by deciding what should consume the image files. A JPEG sequence can be useful for an editor or asset workflow that needs a candidate image. Trick-play imagery belongs with a supported streaming output group, where the player can use the related manifest information. Do not add an image output to a job merely because the word “thumbnail” appears in a delivery brief; first establish whether it means a library poster, a contact sheet, or a seek preview.
In the MediaConvert job, configure the output path for the intended task. For standalone JPEG captures, the output’s video codec is FRAME_CAPTURE, with FrameCaptureSettings for cadence, limit, and quality. For seek previews, choose the image-based trick-play option on the relevant supported streaming output group, then select or configure the mode available for that output. AWS’s frame-capture API reference and image-based trick-play documentation are the appropriate sources to check for current setting names and availability.
Treat job settings as a pair with the input and output timeline. The source may have a different frame rate, timecode basis, or clipping range than the output. AWS explains that MediaConvert assembles output using input and output timelines and supports frame-number timecodes for frame accuracy in its timecode documentation. In practice, a time expressed as ordinary seconds may not name an exact frame; frame-rate conversion, clipping and timecode source can affect where a requested visual moment lands.
Before running a production job, keep a record of the input, the relevant output group, the capture or trick-play settings, and where the result is written. If another person will select the poster image, give them the intended scene or approximate time as well. That avoids treating an arbitrary first capture as the editorial choice.
Configure FRAME_CAPTURE for JPEGs
For standalone captures, add a video output that uses the FRAME_CAPTURE codec and configure FrameCaptureSettings. MediaConvert produces numbered JPEG files rather than a finished poster asset with a human-selected scene. That sequence lets you inspect the frames and decide which JPEG belongs in a catalogue, thumbnail slot, or another downstream use.
The settings are expressed as a rational rate: framerateDenominator divided by framerateNumerator gives the interval in seconds between repeated captures. The first output frame is captured first, before that repeated interval. For example, a numerator of 1 and denominator of 3 captures the first frame and then a frame every three seconds. AWS describes the configuration in the FrameCaptureSettings API reference.
For an approximate poster moment, set a cadence that is likely to include the desired scene, run the job, and inspect the numbered JPEG sequence. If the desired image is around a particular point in a long recording, a wide capture interval could skip it; a tighter interval can produce more candidates but also more files. This is a way to find and select a candidate, not a claim that the console offers an arbitrary timestamp picker that guarantees the exact intended frame.
The file numbering is zero-based. Keep that in mind when matching a capture to a sequence or reviewing a list of filenames. If you need a single deliverable, perform the selection after capture and copy or reference the chosen JPEG in the surrounding workflow. Check the image visually at its intended size: an image that reads clearly when enlarged can be hard to recognise in a small preview.
Set cadence, maximum captures and quality
maxCaptures limits how many JPEG capture files the output produces. Set it with the expected duration and cadence in mind so the job does not create an unnecessarily long sequence. If the cap is too low, the sequence may end before the scene you want; if it is higher than needed, reviewing the output takes longer. The cap bounds the output, but it does not select which image is most suitable as a poster.
quality controls JPEG quality. The API accepts values within its documented bounds; check the current reference when constructing a job rather than relying on a remembered limit. Higher quality can mean larger files, while a lower setting may be adequate for a small preview. Judge the result at the size and use that matter to you, and avoid assuming the same setting is best for every source or destination.
A compact illustrative settings fragment might look like this:
{
"Codec": "FRAME_CAPTURE",
"FrameCaptureSettings": {
"FramerateNumerator": 1,
"FramerateDenominator": 3,
"MaxCaptures": 100,
"Quality": 80
}
}
These numbers illustrate the shape of the settings, not AWS recommendations for every job. The 1 and 3 cadence means the first frame, followed by one capture every three seconds; the maximum and quality should be chosen for the actual output and validated against the current API model. In particular, do not copy a quality bound from an old example without checking the live reference.
When you need a frame tied closely to a known point, think in frames rather than assuming a wall-clock timestamp maps neatly. AWS documents timecodes in HH:MM:SS:FF form for frame-accurate video timecodes; drop-frame notation uses a semicolon before the frame number. The output rate and job timeline still matter. If an exact frame is important, make a test job and verify the extracted image against the intended source moment.
Enable image-based trick-play generation
For seek previews, enable the image-based trick-play option appropriate to the streaming output group. The available modes depend on the output type and the settings interface used. AWS documents modes that can create thumbnails, thumbnails with full frames, or advanced settings; do not assume every format and mode yields the same image types. This is a streaming-oriented output path rather than a shortcut for producing one selected poster.
Advanced settings can expose cadence and tile dimensions. The cadence may follow IDR frames, use a custom interval, or follow segment boundaries. FOLLOW_IFRAME follows IDR cadence; FOLLOW_CUSTOM uses a requested interval; and FOLLOW_SEGMENTATION aligns images to segment boundaries. Segment-following cadence requires 1x1 tiling. Check the corresponding current API reference for the output format, since image dimensions, tile dimensions and divisibility constraints differ across settings and formats.
A custom interval may need to be represented by a compatible output frame boundary. AWS gives the example that at 29.97 frames per second, an entered five-second interval is represented by 150 frames. That illustrates rounding to the output frame rate, not a promise that every requested interval will correspond to an exact real-time duration in every workflow.
IDR-following cadence inherits the output GOP structure. AWS describes typical GOP duration as one to two seconds and notes that streaming segment lengths should be multiples of GOP size in its GOP and segment guidance. Those are encoding considerations, not a guaranteed trick-play interval. If the preview rhythm matters, coordinate GOP and segment choices with your desired cadence, then inspect the generated output in the actual player.
Check output files and manifest references
After the job completes, inspect the destination rather than assuming the output matches the label in the job. With FRAME_CAPTURE, look for the numbered JPEG sequence, confirm it is complete enough for your selection, and open candidate files. Check for the intended visual moment, orientation, framing and legibility at the size where the still will be displayed. If the goal was one poster, explicitly pick one file and keep track of its name.
For trick-play, inspect both the images and the generated manifests. For HLS, AWS documents a child manifest for each image set and corresponding entries in the parent manifest; Roku trick mode is one application described in the HLS trick-play documentation. A manifest entry is useful only if the playback client understands the associated layout and image type. Test seeking with the target player, not just a file browser.
Do not infer from the presence of a tiled image that the player will automatically show it. The manifest, image dimensions, tile layout, and player support all need to line up. Likewise, an output that creates full-frame images in one mode does not prove another mode or format will do so. Keep a small test output and verify the behaviour before using the same settings in a larger delivery workflow.
If the stream itself is a long-running YouTube channel, the media preparation stage is only one part of keeping the programme visible. A practical guide to keeping a devotional stream live without a PC discusses the separate operational question of continuous playback. For the video asset workflow, be sure the selected poster or preview is actually used by the relevant destination; generating a JPEG does not publish it to YouTube automatically.
Choose the method that fits playback
Use standalone captures when you need a simple image sequence, one image selected by an editor, or a poster candidate to hand off to another process. Choose the cadence to cover the desired scene without producing more files than you can review, limit the count, and inspect the numbered result. If only one particular visual moment matters, treat selection as a deliberate downstream step and verify it against the timeline.
Use trick-play image generation when the delivery is a supported streaming output and the viewer needs previews while seeking. Confirm the desired mode exists for that output group, choose a cadence that fits the output’s frame rate and segmenting, and test the manifest in the intended player. The format and mode determine whether the job produces tiles, full-frame images, or a combination; there is no universal image output to assume.
These paths also have different consequences for a small team. JPEG captures are straightforward to review and pass on, but you must choose and integrate the image yourself. Trick-play can connect preview imagery to streaming manifests, but it adds compatibility checks across encoding and playback. For planning a continuous prerecorded YouTube programme, this guide on compressing video for continuous streaming on a slow connection in India covers a separate constraint: the media file and broadcast workflow still need to suit the available connection.
If your remaining problem is that a prepared video must keep playing live after your computer is switched off, StreamNeo removes the need to keep a local machine running the broadcast: you upload the video once and use your YouTube stream key, while the stream is monitored and restarted automatically if it drops.
Before committing, compare the operating options on the pricing page. When the file and channel are ready, start free — 24-hour trial, no card.
FAQ
Does MediaConvert make one poster image automatically?
The documented FRAME_CAPTURE route creates a sequence of numbered JPEGs, not a guaranteed single editorially chosen poster. Select the frame you want from the resulting files. Trick-play output serves seeking previews and may produce different image types depending on format and mode.
How do I extract a frame near a particular time?
Use FRAME_CAPTURE with a cadence likely to include the point, then identify the corresponding numbered JPEG. The capture cadence, input and output frame rates, clipping, and timecode basis affect which frame represents that time. Verify the image rather than treating a wall-clock second as an exact frame address.
Do HLS and DASH use the same thumbnail files?
Not necessarily. Supported output types and modes differ in their settings and image layout, and the playback client must understand the manifest references. Check the current MediaConvert documentation for the chosen output and test it in the target player.
Should I use trick-play images for a YouTube live loop?
Only if the delivery and playback path actually uses a supported streaming manifest and the player supports its image references. A JPEG poster candidate is a separate asset and does not become a YouTube live thumbnail merely because MediaConvert created it. Verify where the still will be used and configure that destination separately.