Watch. Listen. Compare.

MuseTalk Examples: Before and After

Three real MuseTalk results, with their original footage and replacement recordings. Play each clip with sound and compare it yourself.

Compare three MuseTalk examples

MuseTalk changes mouth movement to follow the supplied audio. The footage and recordings come from separate sources; no voice is generated here.

Replacement audio

Video + replacement recording

Natural English Speech · 10 Seconds

Front-facing footage paired with a separate natural English recording. The free-clip result includes a small corner watermark.

Watch for Play all ten seconds. Compare the mouth at speech starts, pauses and the end of the recording.

Use these example files

Run conditions & source credits

fal-ai/musetalk
960 × 540 · 25 fps · continuous video excerpt 0–10s · natural English replacement audio, first 10s · no looped input · free-entitlement run; small delivered corner watermark
A particular MuseTalk run for side-by-side inspection. Review mouth movement and timing yourself; this example does not establish quality, speed or success rate for every input.
Footage: Antoni Shkraba on Pexels, video 8048249 (https://www.pexels.com/video/woman-talking-in-front-of-camera-8048249/), Pexels License (https://www.pexels.com/license/); continuous 0–10s excerpt, resized to 960 × 540 at 25 fps. Audio: LibriSpeech dev-clean, prepared by Vassil Panayotov with Daniel Povey; Panayotov, Chen, Povey and Khudanpur (ICASSP 2015), via OpenSLR SLR12 (https://www.openslr.org/12), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), derived from LibriVox; recording 2277-149896-0032, first 10s, converted to mono 16kHz PCM WAV. Altered-media lip-sync demonstration using separate video and narration; no subject, speaker, creator or dataset endorsement. Provider endpoint: fal-ai/musetalk; exact upstream model revision undisclosed.

Replacement audio · natural male English recording

Video + replacement recording

Office Speaker · Natural Male Voice · 5 Seconds

A man in an office setting, paired with a separate natural male English recording. Listen to the recording, then compare the original and generated mouth movement.

Watch for Compare the lips, teeth and lower face against the original as the speaker moves.

Use these example files

Run conditions & source credits

fal-ai/musetalk
960 × 540 · 25 fps · continuous video excerpt 0–5s · male subject and office background · natural male English replacement audio, LibriSpeech speaker 1272, recording 1272-128104-0000, first 5s · credit-funded QA run; 5 credits captured · no MuseTalk.app product watermark
A real MuseTalk API output from one particular run. The visible speaker, clothing and setting differ from the homepage example. Review mouth movement and timing yourself; this record does not establish quality, speed or success rate for every input.
Footage: Antoni Shkraba on Pexels, video 8048256 (https://www.pexels.com/video/man-talking-in-front-of-camera-8048256/), Pexels License (https://www.pexels.com/license/); continuous 0–5s excerpt, resized to 960 × 540 at 25 fps. Audio: LibriSpeech dev-clean, prepared by Vassil Panayotov with Daniel Povey; Panayotov, Chen, Povey and Khudanpur (ICASSP 2015), via OpenSLR SLR12 (https://www.openslr.org/12/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), derived from LibriVox; recording 1272-128104-0000, first 5s, converted to mono 16kHz PCM WAV. The archive's SPEAKERS.TXT identifies speaker 1272 as male, John Rose. The full archive checksum was not verified; individual downloaded member hashes were recorded. The recorded speaker is separate from the person in the footage. Altered-media lip-sync demonstration; no subject, speaker, creator or dataset endorsement. Provider endpoint: fal-ai/musetalk; exact upstream model revision undisclosed.

Replacement audio · natural French archival recording

Video + replacement recording

Podcast Speaker · Natural French · 5 Seconds

Portrait footage of a woman wearing headphones at a microphone, paired with a separate natural French recording.

Watch for Listen to the supplied recording, then check mouth timing as the speaker gestures and changes head position.

Use these example files

Run conditions & source credits

fal-ai/musetalk
540 × 960 portrait · 25 fps · continuous video excerpt 0–5s · one visible adult speaker with headphones and microphone · natural French replacement audio · credit-funded QA run; 5 credits captured · no MuseTalk.app product watermark
A real MuseTalk API output from one particular run. The portrait framing, visible subject, podcast setting and language differ from the homepage example. Review mouth movement and timing yourself; this record does not establish quality, speed or success rate for every input.
Footage: ANTONI SHKRABA production on Pexels, video 8135208 (https://www.pexels.com/video/8135208/), Pexels License (https://www.pexels.com/license/); continuous 0–5s excerpt, resized to 540 × 960 at 25 fps with source audio removed. Audio: Multilingual LibriSpeech French, by Vineel Pratap, Qiantong Xu, Anuroop Sriram, Gabriel Synnaeve and Ronan Collobert (2020), via OpenSLR SLR94 (https://www.openslr.org/94/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), derived from LibriVox; archive member 11822_11650_000162, first 5s, converted to mono 16kHz PCM WAV. The source member hash was recorded; the full archive checksum was not independently verified. Altered-media lip-sync demonstration using separate video and narration; no subject, speaker, creator or dataset endorsement. Provider endpoint: fal-ai/musetalk; exact upstream model revision undisclosed.

“Use these example files” loads the video and recording into the studio. A new generation starts only after you confirm. See the input and clip selection guide.

What to check in your own clip

Listen for timing through the whole sentence. Inspect the lips, teeth and cheeks during pauses and head movement. These short runs show individual results, not an accuracy or success-rate benchmark. For a delivered clip that looks or sounds wrong, see result troubleshooting.

Earlier runs · 8-second synthetic voice and 5-second English

These MuseTalk examples use the same source subject as the 10-second English clip, with different segments or recordings.

Replacement audio · synthetic demo voice (eSpeak NG)

Verified MuseTalk example

A Clear Front-Facing Speaker

An eight-second silent stock video paired with independently authored synthetic narration. Compare the original mouth movement with the MuseTalk result; the result uses the supplied audio.

The recording is supplied separately from the footage. MuseTalk changes mouth movement; it does not create or improve the voice.

8 seconds · Real MuseTalk output · No sign-in needed to watch

Use these example files

Loads the video and audio into the generator. Nothing is generated until you confirm.

Run conditions & source credits

fal-ai/musetalk
960 × 540 · 25 fps · one visible adult speaker · synthetic replacement narration
A real MuseTalk API output. Review the mouth region and timing yourself; this example does not establish quality for every input.
Footage: Antoni Shkraba on Pexels (8048249), used under the Pexels License. Original narration synthesized with eSpeak NG. Model version beyond the provider endpoint is unverified. Altered-media demonstration; the stock subject and source creator do not endorse MuseTalk.app.

Read captions (WebVTT)

Replacement audio

Verified MuseTalk example

Natural English Speech · 5 Seconds

A continuous 5-second stock-video segment paired with a natural English recording. The supplied recording is separate from the person in the footage; compare the original mouth movement with this run’s MuseTalk result.

5 seconds · Real MuseTalk output · No sign-in needed to watch

Use these example files

Loads the video and audio into the generator. Nothing is generated until you confirm.

Run conditions & source credits

fal-ai/musetalk
960 × 540 · 25 fps · continuous video excerpt 7–12s · natural English replacement audio, first 5s · no looped input · credit-funded QA run; not a cash-purchase demonstration
A particular MuseTalk run for side-by-side inspection. Review mouth movement and timing yourself; this example does not establish quality, speed or success rate for every input.
Footage: Antoni Shkraba on Pexels, video 8048249 (https://www.pexels.com/video/woman-talking-in-front-of-camera-8048249/), Pexels License (https://www.pexels.com/license/); continuous 7–12s excerpt, resized to 960 × 540 at 25 fps. Audio: LibriSpeech dev-clean, prepared by Vassil Panayotov with Daniel Povey; Panayotov, Chen, Povey and Khudanpur (ICASSP 2015), via OpenSLR SLR12 (https://www.openslr.org/12), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), derived from LibriVox; recording 2277-149896-0032, first 5s, converted to mono 16kHz PCM WAV. Altered-media lip-sync demonstration using separate video and narration; no subject, speaker, creator or dataset endorsement. Provider endpoint: fal-ai/musetalk; exact upstream model revision undisclosed.

Try your own video and recording

Start with one clear speaker. Compare the free clip and paid options.

Open the studio

Watch. Listen. Compare.

MuseTalk Examples: Before and After

Three real MuseTalk results, with their original footage and replacement recordings. Play each clip with sound and compare it yourself.

Compare three MuseTalk examples

MuseTalk changes mouth movement to follow the supplied audio. The footage and recordings come from separate sources; no voice is generated here.

Replacement audio

Video + replacement recording

Natural English Speech · 10 Seconds

Front-facing footage paired with a separate natural English recording. The free-clip result includes a small corner watermark.

Watch for Play all ten seconds. Compare the mouth at speech starts, pauses and the end of the recording.

Use these example files

Run conditions & source credits

Replacement audio · natural male English recording

Video + replacement recording

Office Speaker · Natural Male Voice · 5 Seconds

A man in an office setting, paired with a separate natural male English recording. Listen to the recording, then compare the original and generated mouth movement.

Watch for Compare the lips, teeth and lower face against the original as the speaker moves.

Use these example files

Run conditions & source credits

Replacement audio · natural French archival recording

Video + replacement recording

Podcast Speaker · Natural French · 5 Seconds

Portrait footage of a woman wearing headphones at a microphone, paired with a separate natural French recording.

Watch for Listen to the supplied recording, then check mouth timing as the speaker gestures and changes head position.

Use these example files

Run conditions & source credits

“Use these example files” loads the video and recording into the studio. A new generation starts only after you confirm. See the input and clip selection guide.

What to check in your own clip

Earlier runs · 8-second synthetic voice and 5-second English

These MuseTalk examples use the same source subject as the 10-second English clip, with different segments or recordings.

Replacement audio · synthetic demo voice (eSpeak NG)

Verified MuseTalk example

A Clear Front-Facing Speaker

An eight-second silent stock video paired with independently authored synthetic narration. Compare the original mouth movement with the MuseTalk result; the result uses the supplied audio.

The recording is supplied separately from the footage. MuseTalk changes mouth movement; it does not create or improve the voice.

8 seconds · Real MuseTalk output · No sign-in needed to watch

Use these example files

Loads the video and audio into the generator. Nothing is generated until you confirm.

Run conditions & source credits

Read captions (WebVTT)

Replacement audio

Verified MuseTalk example

Natural English Speech · 5 Seconds

5 seconds · Real MuseTalk output · No sign-in needed to watch

Use these example files

Loads the video and audio into the generator. Nothing is generated until you confirm.

Run conditions & source credits

Try your own video and recording

Start with one clear speaker. Compare the free clip and paid options.

Open the studio