Video
video-restyler
Give footage or a playblast a new look.
<!-- Generated by apps/docs/scripts/gen-tools.mjs. Do not edit here: the words live in apps/web/src/lib/tools/video-restyle.tool.json (agent, hint) and packages/shared/src/tools/setting-hints.ts. --> On the rail as **video-restyler**, **flux-3-edit**. ## When to use it Changing a clip you already have: swap or remove an object, relight it, replace the background, or change its style or mood, in words or from a style image. Capture moves a picture along a path you draw; Stop motion redraws every frame. ## When to reach for something else Capture authors motion from a reference image; it is not a general portrait color-editing mode. A clip with pictures of who should be in it goes to Raven, which recasts it. ## How to use it 1. Use the card's **Video & style** area. The clip to change, and an optional style picture. Write the change or let the style picture set it. 2. Set **Draft first** on the card. 3. Pick a model in the pill at the top of the card, or keep **Rothko**. 4. Press **Run**. ## On the card | Control | What it does | Choices | | - | - | - | | Video & style | The clip to change, and an optional style picture. Write the change or let the style picture set it. | | | Draft first | Renders a 480p draft of the new look. Finish the one you like in 1080p, within seven days. | On or off | | Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | | | ×1 | Runs the card several times in a row with the same settings. | 1 to 99 | ## Models The pill at the top of the card picks the model. **Rothko** is the one a new card starts on. | Model | Good for | | - | - | | [Aleph 2](#aleph-2) | Editing a clip you already have from a plain instruction: swap or relight an object, change the background or restyle the look, with the source motion and timing kept, and up to 5 keyframe images pinning an edit to a moment. | | [Kling Omni Std](#kling-omni-std) | Transferring a style image onto a motion video at 720p: the image supplies the subject, the clip supplies the motion and camera, and the output keeps the timing. The lower-cost of the two Kling tiers. | | [Kling Omni Pro](#kling-omni-pro) | The same style-image plus motion-video transfer at 1080p, with higher fidelity than the Standard tier: the image supplies who or what is in the shot, the clip supplies how it moves. | | [Seedance 2.0](#seedance-20) | Restyling a motion clip under a new look from a prompt and an optional style image, with the original motion and camera preserved: 4 to 15 seconds, up to 1080p, optional generated audio. | | [FLUX 3 Edit](#flux-3-edit) | In-place edits of a clip from a text instruction: swap a character, rebuild the background, restyle the shot or rewrite a spoken line with lip sync following it, with motion, timing and framing kept. Billed per second of the source clip. | | [Rothko](#rothko) | Making a playblast, previz or greybox out of 3D into the finished shot from words alone: the camera, the staging and the timing follow the clip, and the words say what each shape becomes and where the shot happens. 4 to 30 seconds. | | [H3 Max](#h3-max) | Turning a Blender or astra playblast into a photoreal shot that keeps its layout, camera and timing, steered by up to eight reference pictures of the characters and places; one continuous shot of 5 to 15 seconds, 480p to 1080p, with sound. | | [LTX 2.3](#ltx-23) | Turning a viewport or hardware render from your 3D software into a photoreal clip with the same motion and framing, with a Light or Strong intensity and an optional still beauty render pinned as the first frame. | | [Ray 3.2](#ray-32) | Capture on Video Restyle: drop a still, drag a circle across it to draw a 3 to 10 second motion path, and the moving shape becomes the subject you name, with a strength dial from adhere to reimagine. | | [Stop motion](#stop-motion) | Stop motion: footage redrawn as hand-made animation, every frame repainted in a look described in words, per material, while the composition and the motion stay exactly as filmed. | ### Aleph 2 **Not the one for:** The source clip must be 2 to 30 seconds and under 16MB, and the result keeps its length and shape; it edits footage rather than animating a still, so it needs a video to start from. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | A number | ### Kling Omni Std **Not the one for:** Needs both a style image and a 3 to 10 second motion video; duration and aspect follow the video, and output tops out at 720p, so a 1080p deliverable wants the Pro tier. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Negative prompt | List what you do NOT want to see, like blurry or extra limbs. | Your words | | Keep original sound | Carries over the original video’s sound. | On or off | ### Kling Omni Pro **Not the one for:** Needs both a style image and a 3 to 10 second motion video and costs more per second than the Standard tier; it cannot edit a clip from text alone. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Negative prompt | List what you do NOT want to see, like blurry or extra limbs. | Your words | | Keep original sound | Carries over the original video’s sound. | On or off | ### Seedance 2.0 **Not the one for:** The motion clip is required and must be 480p to 720p and under 50MB; reference-clip seconds are billed on top of the output length, and there is no strength dial. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 480p, 720p, 1080p | | Generate audio | Generate synchronized audio in the same pass. | On or off | ### FLUX 3 Edit **Not the one for:** The clip must be an MP4 under 15 seconds and the result is exactly as long as the input; it cannot lengthen a clip and takes no style image. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Safety tolerance | How permissive the content filter is. 0 is strictest, 4 is most permissive. | 0 to 4 | ### Rothko **Not the one for:** It takes no pictures: pictures of the look go to H3 Max on the same source, and pictures of who should be in the shot go to plnty-raven. It needs a few words to run. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Resolution | 1080p adds a finishing pass for detail and takes longer. | 480p, 720p, 1080p | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | A number | ### H3 Max **Not the one for:** It needs one continuous shot: a clip with cuts fails. The camera and the motion come from the playblast, so words cannot change them, and it stops at 15 seconds. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Resolution | Each step up looks sharper and costs more per second. | 480p, 768p, 1080p | ### LTX 2.3 **Not the one for:** Made for CG renders, not for restyling live footage; output is 480p or 720p, and the bill follows frames (duration times fps), so the frame rate is the cost dial. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Intensity | How hard to push toward photoreal. Light keeps the render's look, strong transforms it more. | Light, Strong | | Resolution | Output resolution (short side). Higher looks sharper and costs more credits. | 480p, 720p | | FPS | Playback speed of the output. Also sets how many frames the clip becomes, so it scales the credit cost. | 8 to 60 | | Generate audio | Add a generated audio track to the output video. | On or off | | Steps | Quality vs speed. More steps look better but take longer. | 15 to 30 | | Guidance | How strictly the model follows your prompt. Higher sticks closer to the text. | 0 to 10 | | Negative prompt | Things to steer away from. Leave empty to use the model default. | Your words | | Expand prompt | Let the model expand your prompt before generating. The 3DREAL trigger is always kept. | On or off | | Safety checker | Run the content safety check on the result. | On or off | | Video quality | Encode quality of the output file. | Low, Medium, High, Maximum | | Write mode | Trade encode speed against file size. | Fast, Balanced, Small file | | Seed | Fix the random seed to reproduce a result. Leave empty for a new random each run. | A number | ### Ray 3.2 **Not the one for:** Capture only: it does not restyle a clip you already have, which is 2D/Footage or Playblast on the same card; HDR doubles the cost. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Strength | How far to reinvent the clip: Auto lets the model decide; adhere keeps it close, flex balances, reimagine lets the prompt drive. Ignored when any conditioning signal is on. | Auto, Adhere 1, Adhere 2, Adhere 3, Flex 1, Flex 2, Flex 3, Reimagine 1, Reimagine 2, Reimagine 3 | | Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 540p, 720p, 1080p | | HDR | Cinema-grade HDR color. Needs 720p or 1080p, and doubles the credit cost. | On or off | | Pose conditioning | Lock the output to the source clip's motion and pose. Overrides the strength preset. | On or off | | Pose detail | How tightly pose tracks the source: precise is strict, coarse is looser. | Precise, Coarse | | Face conditioning | Preserve facial identity from the source clip. Overrides the strength preset. | On or off | | Depth conditioning | Hold the source scene's depth and layout. Overrides the strength preset. | On or off | | Depth blur | Soften the depth signal. 0 keeps the model default; higher blurs it. | 0 to 1 | | Normals conditioning | Condition on surface normals (shading direction). Overrides the strength preset. | On or off | | Normals detail | Strength of the normals signal. 0 keeps the model default. | 0 to 1 | | Trajectory conditioning | Condition on motion trajectory. Overrides the strength preset. | On or off | | Trajectory sparsity | How sparse the motion sampling is. 0 keeps the model default. | 0 to 1 | ### Stop motion **Not the one for:** It takes the look in words only, never a style picture, which would drag the motion toward its own pose. Cadence changes the texture of the wobble, not the length or the camera move. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Cadence | The animation's frame rate. On twos (12 fps) is the classic hand-drawn stop-motion feel; on ones (24) is smoother, on threes (8) is choppier. The output always plays the full length of your clip. | On ones, On twos, On threes | | Quality | High paints 4 frames per generation instead of 8, so every frame comes out larger and sharper. Costs about 2x the credits. | Standard, High |On the rail as video-restyler, flux-3-edit.
When to use it
Section titled “When to use it”Changing a clip you already have: swap or remove an object, relight it, replace the background, or change its style or mood, in words or from a style image. Capture moves a picture along a path you draw; Stop motion redraws every frame.
When to reach for something else
Section titled “When to reach for something else”Capture authors motion from a reference image; it is not a general portrait color-editing mode. A clip with pictures of who should be in it goes to Raven, which recasts it.
How to use it
Section titled “How to use it”- Use the card’s Video & style area. The clip to change, and an optional style picture. Write the change or let the style picture set it.
- Set Draft first on the card.
- Pick a model in the pill at the top of the card, or keep Rothko.
- Press Run.
On the card
Section titled “On the card”| Control | What it does | Choices |
|---|---|---|
| Video & style | The clip to change, and an optional style picture. Write the change or let the style picture set it. | |
| Draft first | Renders a 480p draft of the new look. Finish the one you like in 1080p, within seven days. | On or off |
| Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | |
| ×1 | Runs the card several times in a row with the same settings. | 1 to 99 |
Models
Section titled “Models”The pill at the top of the card picks the model. Rothko is the one a new card starts on.
| Model | Good for |
|---|---|
| Aleph 2 | Editing a clip you already have from a plain instruction: swap or relight an object, change the background or restyle the look, with the source motion and timing kept, and up to 5 keyframe images pinning an edit to a moment. |
| Kling Omni Std | Transferring a style image onto a motion video at 720p: the image supplies the subject, the clip supplies the motion and camera, and the output keeps the timing. The lower-cost of the two Kling tiers. |
| Kling Omni Pro | The same style-image plus motion-video transfer at 1080p, with higher fidelity than the Standard tier: the image supplies who or what is in the shot, the clip supplies how it moves. |
| Seedance 2.0 | Restyling a motion clip under a new look from a prompt and an optional style image, with the original motion and camera preserved: 4 to 15 seconds, up to 1080p, optional generated audio. |
| FLUX 3 Edit | In-place edits of a clip from a text instruction: swap a character, rebuild the background, restyle the shot or rewrite a spoken line with lip sync following it, with motion, timing and framing kept. Billed per second of the source clip. |
| Rothko | Making a playblast, previz or greybox out of 3D into the finished shot from words alone: the camera, the staging and the timing follow the clip, and the words say what each shape becomes and where the shot happens. 4 to 30 seconds. |
| H3 Max | Turning a Blender or astra playblast into a photoreal shot that keeps its layout, camera and timing, steered by up to eight reference pictures of the characters and places; one continuous shot of 5 to 15 seconds, 480p to 1080p, with sound. |
| LTX 2.3 | Turning a viewport or hardware render from your 3D software into a photoreal clip with the same motion and framing, with a Light or Strong intensity and an optional still beauty render pinned as the first frame. |
| Ray 3.2 | Capture on Video Restyle: drop a still, drag a circle across it to draw a 3 to 10 second motion path, and the moving shape becomes the subject you name, with a strength dial from adhere to reimagine. |
| Stop motion | Stop motion: footage redrawn as hand-made animation, every frame repainted in a look described in words, per material, while the composition and the motion stay exactly as filmed. |
Aleph 2
Section titled “Aleph 2”Not the one for: The source clip must be 2 to 30 seconds and under 16MB, and the result keeps its length and shape; it edits footage rather than animating a still, so it needs a video to start from.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | A number |
Kling Omni Std
Section titled “Kling Omni Std”Not the one for: Needs both a style image and a 3 to 10 second motion video; duration and aspect follow the video, and output tops out at 720p, so a 1080p deliverable wants the Pro tier.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Negative prompt | List what you do NOT want to see, like blurry or extra limbs. | Your words |
| Keep original sound | Carries over the original video’s sound. | On or off |
Kling Omni Pro
Section titled “Kling Omni Pro”Not the one for: Needs both a style image and a 3 to 10 second motion video and costs more per second than the Standard tier; it cannot edit a clip from text alone.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Negative prompt | List what you do NOT want to see, like blurry or extra limbs. | Your words |
| Keep original sound | Carries over the original video’s sound. | On or off |
Seedance 2.0
Section titled “Seedance 2.0”Not the one for: The motion clip is required and must be 480p to 720p and under 50MB; reference-clip seconds are billed on top of the output length, and there is no strength dial.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 480p, 720p, 1080p |
| Generate audio | Generate synchronized audio in the same pass. | On or off |
FLUX 3 Edit
Section titled “FLUX 3 Edit”Not the one for: The clip must be an MP4 under 15 seconds and the result is exactly as long as the input; it cannot lengthen a clip and takes no style image.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Safety tolerance | How permissive the content filter is. 0 is strictest, 4 is most permissive. | 0 to 4 |
Rothko
Section titled “Rothko”Not the one for: It takes no pictures: pictures of the look go to H3 Max on the same source, and pictures of who should be in the shot go to plnty-raven. It needs a few words to run.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Resolution | 1080p adds a finishing pass for detail and takes longer. | 480p, 720p, 1080p |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | A number |
H3 Max
Section titled “H3 Max”Not the one for: It needs one continuous shot: a clip with cuts fails. The camera and the motion come from the playblast, so words cannot change them, and it stops at 15 seconds.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Resolution | Each step up looks sharper and costs more per second. | 480p, 768p, 1080p |
LTX 2.3
Section titled “LTX 2.3”Not the one for: Made for CG renders, not for restyling live footage; output is 480p or 720p, and the bill follows frames (duration times fps), so the frame rate is the cost dial.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Intensity | How hard to push toward photoreal. Light keeps the render’s look, strong transforms it more. | Light, Strong |
| Resolution | Output resolution (short side). Higher looks sharper and costs more credits. | 480p, 720p |
| FPS | Playback speed of the output. Also sets how many frames the clip becomes, so it scales the credit cost. | 8 to 60 |
| Generate audio | Add a generated audio track to the output video. | On or off |
| Steps | Quality vs speed. More steps look better but take longer. | 15 to 30 |
| Guidance | How strictly the model follows your prompt. Higher sticks closer to the text. | 0 to 10 |
| Negative prompt | Things to steer away from. Leave empty to use the model default. | Your words |
| Expand prompt | Let the model expand your prompt before generating. The 3DREAL trigger is always kept. | On or off |
| Safety checker | Run the content safety check on the result. | On or off |
| Video quality | Encode quality of the output file. | Low, Medium, High, Maximum |
| Write mode | Trade encode speed against file size. | Fast, Balanced, Small file |
| Seed | Fix the random seed to reproduce a result. Leave empty for a new random each run. | A number |
Ray 3.2
Section titled “Ray 3.2”Not the one for: Capture only: it does not restyle a clip you already have, which is 2D/Footage or Playblast on the same card; HDR doubles the cost.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Strength | How far to reinvent the clip: Auto lets the model decide; adhere keeps it close, flex balances, reimagine lets the prompt drive. Ignored when any conditioning signal is on. | Auto, Adhere 1, Adhere 2, Adhere 3, Flex 1, Flex 2, Flex 3, Reimagine 1, Reimagine 2, Reimagine 3 |
| Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 540p, 720p, 1080p |
| HDR | Cinema-grade HDR color. Needs 720p or 1080p, and doubles the credit cost. | On or off |
| Pose conditioning | Lock the output to the source clip’s motion and pose. Overrides the strength preset. | On or off |
| Pose detail | How tightly pose tracks the source: precise is strict, coarse is looser. | Precise, Coarse |
| Face conditioning | Preserve facial identity from the source clip. Overrides the strength preset. | On or off |
| Depth conditioning | Hold the source scene’s depth and layout. Overrides the strength preset. | On or off |
| Depth blur | Soften the depth signal. 0 keeps the model default; higher blurs it. | 0 to 1 |
| Normals conditioning | Condition on surface normals (shading direction). Overrides the strength preset. | On or off |
| Normals detail | Strength of the normals signal. 0 keeps the model default. | 0 to 1 |
| Trajectory conditioning | Condition on motion trajectory. Overrides the strength preset. | On or off |
| Trajectory sparsity | How sparse the motion sampling is. 0 keeps the model default. | 0 to 1 |
Stop motion
Section titled “Stop motion”Not the one for: It takes the look in words only, never a style picture, which would drag the motion toward its own pose. Cadence changes the texture of the wobble, not the length or the camera move.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Cadence | The animation’s frame rate. On twos (12 fps) is the classic hand-drawn stop-motion feel; on ones (24) is smoother, on threes (8) is choppier. The output always plays the full length of your clip. | On ones, On twos, On threes |
| Quality | High paints 4 frames per generation instead of 8, so every frame comes out larger and sharper. Costs about 2x the credits. | Standard, High |