Your clip plays back wrong: limbs snap between frames, the subject teleports a few pixels every fifth frame, body parts judder at the moment one motion hands off to another. That is motion jitter, and it is a different failure from background flicker (environment instability) and drift (identity loss). Jitter is the motion path itself failing to interpolate smoothly.
Fastest fix (works in most cases): cut your prompt down to one primary action, anchor the start pose with image-to-video, and render at the tool’s native frame rate (Kling 3.0 at 30 fps with moderate Motion Intensity, Runway Gen-4.5 with one Motion Brush vector, or Veo 3.1 which is the most jitter-resistant out of the box as of June 2026). If you only change one thing, drop the second and third action verbs from your prompt. That alone clears most jitter.
Which bucket are you in
Use this to point yourself at the right section instead of trying everything.
| What you see | Most likely cause | Jump to |
|---|---|---|
| Limbs snap when one move becomes another | Multiple actions in one clip | Cause 1 / Step 1 |
| First half-second jerks, then settles | Start pose unspecified | Cause 2 / Step 2 |
| Constant low-grade shimmer in fast motion | Generated at low fps, interpolated up | Cause 3 / Step 3 |
| Subject “teleports” into a new pose | End pose implies a discontinuity | Cause 4 |
| One subject smooth, the other juddery | Multiple independent motion paths | Cause 5 / Step 7 |
| Subject jitters during a tracking shot | Camera speed matches subject speed | Cause 6 |
| Calm scene that should be still keeps twitching | Motion intensity set too high | Cause 7 / Step 4 |
Common causes
Ordered by what trips creators most often.
1. Multiple actions stuffed into one prompt
Running while waving, then turning to look back, then jumping is three distinct actions in a 3-second clip. The model tries to fit all of them and produces a jerky path that snaps between them with no in-between frames.
How to spot it: count the action verbs in your prompt. More than one primary action per 3 seconds and you are in jitter territory.
2. Start pose unspecified
The model picks a random starting pose, then animates toward whatever the prompt implies. The random start often does not align with the prompted action, producing a snap correction in the first frames.
How to spot it: compare frame 1 with frame 5. If frame 1’s pose makes the prompted action impossible (subject facing the wrong way, wrong limb position), the model is fighting its own start.
3. Low generation fps interpolated up
Some pipelines generate at 12-16 fps internally, then interpolate up to 24 or 30 fps for delivery. Interpolation does not invent missing motion, it cross-fades between the frames it has. Result: judder, worst at fast-motion moments.
How to spot it: check the tool’s delivery fps against its native generation fps. As of June 2026: Runway Gen-4.5 delivers 24 or 25 fps (selectable in Advanced settings); Kling 3.0 supports up to 60 fps natively (30 fps is a low-jitter filmic default); Pika 2.5 delivers ~5-second clips at 24 fps; Google Veo 3.1 outputs up to 4K at 60 fps after its January 2026 update, with 1080p available at 24, 30, or 60 fps and the strongest temporal coherence in this group. If a 24 fps file came from a low-fps internal pass, jitter is baked in.
4. End pose implies a discontinuity
Standing still, then suddenly running forces the model to invent a transition. It often teleports the limbs straight into the running pose rather than accelerating smoothly into it.
5. Multiple subjects with independent motion
Two characters dancing, two cars racing: each subject has its own motion path, and the model often jitters one to keep the other consistent.
6. Camera motion at the same speed as subject motion
A tracking shot moving at the subject’s speed makes the subject look static while the background flies by. The model then jitters the subject to “show” it is moving.
7. Motion intensity too high for the subject
A clip of a person calmly drinking coffee belongs at low motion. Kling’s Motion Intensity slider (shown as a 0.1-1.0 range on some surfaces and a 1-10 range on others) and Pika’s motion-strength slider often sit high enough by default that static-to-mild shots pick up unwanted jitter.
Before you change anything
- Save the prompt, the model and version, and the jittery output clip.
- Note whether jitter is at hand-off moments (start, middle, end) or continuous throughout.
- Record the model, the motion settings, and the clip length.
- Decide whether the clip truly needs the full action sequence, or whether a simpler action would serve.
- Back up the prompt template before editing it.
Information to collect
- Full prompt, motion intensity/strength, model, version.
- Clip length and aspect ratio.
- Specific timestamps where jitter is worst.
- Whether the jitter pattern is consistent across multiple regenerations (a structural prompt issue) or varies run to run (closer to random sampling).
- The intended motion type (calm, moderate, fast).
Shortest path to fix
Step 1: One primary action per clip
Rewrite the prompt around a single verb.
Before:
a woman running through a city, looking back over her shoulder,
then jumping over a fence, motion blurAfter:
a woman running through a city at moderate speed, slight head turn,
smooth continuous motionIf the use case truly needs multiple actions, split it into separate clips and cut between them. Kling 3.0’s multi-shot mode (3-15 second sequences with subject consistency across angles) is built for exactly this — let it cut, rather than cramming the actions into one shot.
Step 2: Specify the start pose
Add an explicit starting pose to anchor frame 1:
starting in a standing pose, then beginning to run, smooth acceleration
from still to motionFor image-to-video, the reference image is your start pose. Choose a reference where the pose makes the upcoming motion physically plausible (for a running clip, pick an image where the subject already leans forward).
Step 3: Render at the tool’s native fps
As of June 2026:
- Runway Gen-4.5: 24 or 25 fps, chosen in Advanced settings
- Kling 3.0: up to 60 fps native; 30 fps is the filmic sweet spot for low jitter
- Pika 2.5: ~5-second clips at 24 fps (extendable via Pikaframes)
- Google Veo 3.1: up to 4K at 60 fps since the January 2026 update; at 1080p you can pick 24, 30, or 60 fps. Strongest temporal coherence in this group. For visibly smoother fast motion, render at 60 fps and time-remap in post if you need a cinematic look.
Pick the model and tier that deliver natively at your target fps instead of a cheaper path that interpolates up. Note: OpenAI shut down the Sora 2 consumer web and app on April 26, 2026 (the API follows on September 24, 2026), so it is no longer a live option for new consumer work. Veo 3.1 and Kling 3.0 are the smooth-motion replacements.
Step 4: Set motion intensity to match the action speed
Match the slider to the real speed of the action, then reinforce it in the prompt. Kling’s Motion Intensity appears as a 0.1-1.0 slider on some surfaces and a 1-10 slider on others, so think in terms of the lower, middle, and upper thirds of whatever scale you see: keep it in the middle for natural-looking movement and only push toward the top for genuine sprints or athletic action. If a generation shows artifacts, drop intensity first before changing anything else. It is the single most impactful parameter in Kling image-to-video.
For slow / calm motion (set Motion Intensity in the lower third of the slider):
slow continuous movement, fluid motion, soft transitionsFor fast / athletic motion (intensity in the upper third):
sharp athletic motion, motion blur, dynamic accelerationThe descriptor and the slider together tell the model how much temporal smoothing to apply. A calm scene with the slider cranked high is a classic jitter source.
Step 5: Use image-to-video for precise control
Text-to-video lets the model start anywhere. Image-to-video pins frame 1 to your reference. For any clip where the starting pose matters, use image-to-video. Kling reports image-to-video as its strongest mode for spatio-temporal stability, and its recent models cut warping, morphing, and flicker by roughly 60% versus the prior generation.
Step 6: Add explicit smooth-motion cues, and keep them physical
smooth motion, no jitter, continuous frame-to-frame coherence,
no pose discontinuityThis is a soft hint but measurably helps on several models. On Runway Gen-4.5 and Veo 3.1’s physics-aware engine, physical force language works even better than negative cues. Describe momentum and weight (“each step presses deep into the snow”, “steady dolly in, no shake”) rather than only saying “no jitter”.
Step 7: For multi-subject scenes, separate and composite
Render each subject in its own clip over a simple background, then composite in post. Asking the model to coordinate two independent motion paths in one generation is the highest-risk approach to jitter.
Step 8: If you use a Motion Brush, make the brush and the prompt agree
In Runway’s Motion Brush, the painted motion vectors must match the text prompt. If you brush a subject moving left but the prompt says it stands still, the model tears and jitters trying to satisfy both. Paint foreground and background with different speeds (faster foreground, slower background) for clean depth instead of competing instructions.
How to confirm the fix
- Play at 25% speed: limbs should follow continuous paths, not snap.
- Pause every 6th frame: the pose should evolve smoothly, not skip.
- Generate three clips at the same settings: all three should be smooth, not just one lucky output.
- Have someone else watch it: they should not flinch at any motion moment.
If it still fails
- Reduce to a single second of motion. If even 1 second jitters, the prompt has a structural issue (go back to Step 1).
- Switch to image-to-video with a strong, motion-plausible reference pose.
- Try a different model. Kling 3.0 leads on motion coherence and low jitter (and now renders up to 60 fps natively); Veo 3.1 is strongest on natural human motion (and can render 60 fps for the smoothest fast action); Runway Gen-4.5 is best when you need frame-level creative control via Motion Brush.
- Render at the slowest speed your prompt allows, then time-remap (speed up) in post if you need faster motion.
- Package the prompt, model and version, output, and jitter timestamps before asking for community help.
FAQ
Is jitter the same as background flicker? No. Jitter is the subject’s motion path failing to interpolate (snapping limbs, teleporting). Flicker is the background or lighting changing frame to frame. They have different fixes — see the background-flicker guide linked below.
Which model has the least jitter as of June 2026? For smooth human motion, Veo 3.1 (up to 4K at 60 fps, strong temporal coherence). For low jitter on action sequences, Kling 3.0 (up to 60 fps native, with reported ~60% fewer warp/morph/flicker artifacts than the previous generation). Runway Gen-4.5 trades a little raw smoothness for the most precise frame-level control via Motion Brush.
Should I just run my jittery clip through a frame-interpolation tool? Interpolation (e.g. converting 30 to 60 fps) can soften mild judder, but it cannot fix a snap caused by a missing in-between pose — it cross-fades and often smears the snap instead. Fix the cause at generation time first; use interpolation only as a final polish.
Will adding “no jitter, smooth motion” to the prompt fix it? It helps a little, but it is a soft cue. The structural fixes (one action per clip, anchored start pose, native fps, matched motion intensity) do the real work. On Runway Gen-4.5 and Veo 3.1, physical force language outperforms negative cues.
Can I use Sora to fix this? No. OpenAI shut down the Sora 2 consumer web and app on April 26, 2026, and the Sora 2 API is scheduled to follow on September 24, 2026, so it is not a live option for new consumer work. Use Veo 3.1 or Kling 3.0 instead.
Prevention
- Storyboard before generating; one action per clip is the rule.
- Plan multi-action sequences as several clips concatenated (or Kling’s multi-shot mode), never one shot.
- Default to image-to-video for any clip where the start pose matters.
- Standardize on one model’s native fps across a project; do not mix frame rates.
- Set the motion-intensity slider to the action’s real speed, and use motion descriptors that match it.