Skip to main content

MeiGen-AI’s InfiniteTalk is a standard open-weight model under Apache-2.0 for sparse-frame video dubbing, regenerating speech-driven expressions and posture with speaker masking and 480p/720p output.

From $0.036/s

Workbench

Input

Lip sync
155/2000
Optional lip-sync direction for expression, action, and lighting, up to 2,000 characters. May be omitted; dialogue comes from audio. Wording does not change the audio-and-resolution charge. at least 1 characters · at most 2000 characters
1/1

Drop a video here, or browse · 1 file · Max 90 MB · MP4 · WEBM

  • 1
1 / 1 files
Required footage for lip-sync with the intended person clearly visible. No default applies; video duration does not set the billed seconds, which follow audio and resolution. at least 1 characters
affects price1/1
1 / 1 files
Required replacement recording for lip-sync, 3–600 seconds with no default. Sets output length and billed duration: round up to whole seconds, minimum three seconds, at the chosen resolution rate. at least 1 characters
0/1

Drop files or media links, paste, or browse · 1 file · Max 10 MB · JPG · PNG · WEBP

0 / 1 files
Optional speaker-selection image for lip-sync when several people appear; mark the intended person white. Omit for an unmasked request. Masking adds no separate charge. at least 1 characters
affects price
Video lip-sync accepts 480p or 720p and defaults to 480p. Every billed second at 720p is twice the 480p rate; resolution does not change which input duration is counted.
Seed
Optional random seed for video lip-sync comparisons; no numeric range or default is specified here. Hold it fixed while changing one setting, without expecting identical output. No separate price effect. no lower than 0 · no higher than 2147483647
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result

Missing the model you need?

AUDIO-DRIVEN VIDEO WITH SPEAKER SELECTION

InfiniteTalk API at a glance

Start with a portrait or footage; let your recording guide the speech, expression, and movement.

Output tiers
2
The 720p per-second rate is twice the 480p rate.
Recording range
3–600 S
InfiniteTalk accepts audio from three seconds through 10 minutes.
Speaker selection
MASK
An optional mask marks the intended speaker in a photo containing several people.
Controlled comparisons
SEED
Reuse the random source while changing one setting; matching output is not guaranteed.
Model license
Apache
MeiGen-AI makes InfiniteTalk available as an open-source model.

Results

What you can make with InfiniteTalk

  • ROOFTOP ASTRONOMY

    A short explanation of Saturn

    In our tests, the astronomer spoke beside his rooftop telescope for about nine seconds. The delivery followed the recording with subtle natural movement.

    Our 480p astronomy sample follows a roughly nine-second recording about Saturn with subtle natural movement.

  • LATE-NIGHT DIALOGUE

    A bar portrait delivers the line

    Our 480p bar example animates a man with a loosened tie saying, 'Last call already? Then stay for one more.' The result includes natural head movement against the warm bar setting.

    Our 480p sample brings the bar portrait to speech with natural head movement as he says, 'Last call already? Then stay for one more.'

  • SPEAKER AND COMPARISON CONTROLS

    Identify the person, then vary one detail

    Mark the intended speaker white in an optional mask when several people share the photo. A fixed seed helps compare one prompt or resolution adjustment at a time, without promising an identical result.

Content test

Is InfiniteTalk uncensored? What we tested

Content levels tested on InfiniteTalk, by mode
LevelLip syncTalking avatar
Mild
Not tested yet
Not tested yet
Moderate
RendersTested Nov 15, 2026
RendersTested Nov 15, 2026
Strong
Not tested yet
Not tested yet
Very strong
Not tested yet
Not tested yet
Maximum
Not tested yet
Not tested yet
How we test

How each mode behaves

Talking avatar
For InfiniteTalk, use photos and recordings you have permission to use; do not impersonate the people depicted or recorded.
Talking avatar
InfiniteTalk audio must be at least 3 seconds and no longer than 10 minutes; the minimum billable duration is 3 seconds.
All modes
InfiniteTalk output files remain available for 14 days before deletion.

Use cases

What to build with the InfiniteTalk API

  • Give a still character a recorded reply

    Use talking-avatar when your starting material is a portrait. Your recording supplies the voice and line, while the model adds speaking motion.

  • Change the mood of existing footage

    Choose lip-sync for a new delivery from a source clip. In our flower-workshop test, a hesitant-to-delighted recording brought a raised head, brighter expression, and shoulder movement.

  • Build a presenter clip from narration

    Pair a presenter photo or video with a prepared explanation. Audio from 3 to 600 seconds is accepted; review the full result when using a longer recording.

  • Direct one person in a shared scene

    Mark the intended speaker white in a mask supplied through mask_url. This selects a person in the source media; the displayed florist test uses a single speaker without a mask.

  • Refine a delivery before a longer run

    Try a short recording and a restrained motion prompt at 480p. Compare one changed instruction with a fixed seed, then decide whether to use 720p for the final clip at twice the rate.

Prompts

InfiniteTalk prompt examples you can copy

  • Talking avatar

    A quiet bar invitation

    Result from this prompt

    A man at a late-night bar speaks softly to someone across from him, a slow half smile, warm amber light.

    Why it works. Keeps the direction centered on a speaking performance in the supplied bar setting; the actual line belongs in the audio recording.

    Try this prompt
  • Talking avatar

    Saturn beside the telescope

    Result from this prompt

    A man talks warmly about the night sky next to his telescope, natural head movement.

    Why it works. Uses the telescope setting to frame a restrained explanation, following the kind of motion seen in our astronomy sample.

    Try this prompt
  • Talking avatar

    A gallery welcome

    A curator in her fifties in a charcoal suit welcomes the viewer from a quiet gallery, speaking with an attentive expression and a slight nod under diffuse exhibition lights.

    Why it works. Offers one clear speaking action and a modest motion cue for an adult presenter portrait.

    Try this prompt
  • Talking avatar

    A candlelit dinner reply

    A woman in her late thirties in a satin evening dress speaks to her partner across a dinner table, her smile turning playful with a small head tilt in candlelight.

    Why it works. Gives a close conversation a mood and movement cue; use a white speaker mask if the source photo includes her partner.

    Try this prompt
  • Talking avatar

    A relaxed studio introduction

    A man in his forties in an open-collar linen shirt introduces his work from a daylight studio, speaking with an easy smile and small, unhurried head movements.

    Why it works. Connects wardrobe, setting and delivery without loading the instruction with a chain of actions.

    Try this prompt
  • Talking avatar

    A morning conversation

    A woman in her forties wearing a silk robe talks to her partner by a sunlit window, a gentle smile and a small nod giving the moment an affectionate mood.

    Why it works. Describes a scene already present in the input portrait and leaves the spoken message to the recording; this is a suggested direction, not a measured sample.

    Try this prompt

Pricing

InfiniteTalk API pricing

A 5-second clip starts at $0.18 at 480p and $0.36 at 720p.

Get started

How to use the InfiniteTalk API in three steps

  1. Step 1

    Choose the visual starting point

    Select talking-avatar and image_url for a photo, or lip-sync and video_url for footage. Check that you have rights to the face and recording; every person shown must be an adult, with no impersonation.

  2. Step 2

    Listen to the line and mark the speaker

    Supply audio_url with a 3–600-second recording of the desired words. If several people appear, a mask_url image can identify the speaker with a white region.

  3. Step 3

    Set motion and the resolution rate

    Describe expression or movement in the optional prompt and select 480p or 720p. The video default is 480p; 720p doubles the per-second rate, and a fixed seed helps compare a single revision.

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

FAQ

InfiniteTalk API FAQ

About InfiniteTalk

Build an InfiniteTalk clip around the recording you want the portrait to deliver, then use a short prompt to guide the scene and movement. Our rooftop astronomy sample showed synchronized speech with subtle head motion, while the bar portrait delivered a late-night line at 480p. Choose 720p when that output tier suits your project, accounting for twice the per-second rate. For a group photo, mark the intended speaker in white. Review the result against your audio and save the finished file within its 14-day availability window.

Related models and next tasks