AI Video Editor Talking Head API Documentation

AI Video Editor Talking Head API Documentation

Playground

AI Video Editor Talking Head cleans up a talking-to-camera video: it removes filler words, false starts, repeats and dead air, keeps the speaker framed, and burns in accurate word-timed captions. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Talking Head automatically cleans up a single talking-to-camera recording while preserving the speaker’s intended message. It removes filler words, false starts, repeated takes, and unnecessary pauses, keeps the speaker framed, and adds verified word-timed captions to produce a tighter, more polished version of the original recording.

Use Talking Head when you want to keep the content of a spoken video but make the delivery cleaner and more concise. It is designed for creator videos, presentations, tutorials, interviews, product explanations, courses, and other speaker-focused footage where the goal is to remove mistakes and dead air rather than create a new highlight from selected moments.

Why Choose This?

• Automatic talking-head cleanup
  Remove filler words, restarts, repeated takes, and dead air without manually editing the timeline.

• Natural sentence-level cuts
  Speech edits are made around sentence boundaries to help keep transitions natural and avoid abrupt audio cuts.

• Captions under your control
  Word-by-word highlighted captions by default, plain static captions, or none at all with caption_style. Captions are added only where the transcription is verified, so nothing is captioned with words that were not clearly said.

• Speaker-aware reframing
  When changing aspect ratio, the frame follows the speaker’s face while slides and title cards without a visible face remain fully shown.

• Multiple output ratios
  Create landscape, vertical, square, portrait, or ultrawide edits for different publishing platforms.

• Long-recording support
  Process up to 120 minutes of source video in a single request.

Parameters

How to Use

1. Upload your recording — Provide the talking-head video you want to clean up.
2. Choose a caption style (optional) — Use caption_style to keep word-by-word highlighting, switch to plain captions, or turn captions off.
3. Choose aspect ratio (optional) — Select a vertical, square, landscape, or other supported format, or leave it unset to follow the source.
4. Choose language (optional) — Specify the spoken language or keep auto for automatic detection.
5. Submit — Let the editor analyze the speech, remove unwanted sections, reframe the speaker, and generate verified captions.
6. Retrieve the result — Download the completed edited and captioned video.

Pricing

Pricing is based on the source video duration.

The rate is $0.005 per 5 seconds, equivalent to $0.06 per minute.

Source duration is rounded up to the next whole second, with a minimum billed duration of 3 seconds and a maximum billed duration of 7200 seconds (120 minutes).

Only the processed source duration affects pricing. aspect_ratio, caption_style and language add no separate charge.

Best Use Cases

• Talking-head videos — Clean up direct-to-camera recordings automatically.
• Creator content — Tighten YouTube, short-form, educational, or commentary videos.
• Interviews — Remove pauses, restarts, and repeated takes from speaker-focused footage.
• Tutorials and explainers — Improve pacing while preserving the speaker’s intended content.
• Presentations — Turn long spoken recordings into cleaner, more polished videos.
• Vertical social content — Reframe landscape recordings for 9:16, 4:5, or square feeds.
• Captioned speech content — Generate burned-in captions with verified word timing.

Pro Tips

• Use clean recordings with clearly audible speech for stronger transcription and edit decisions.
• Set language explicitly when you know the spoken language and want more predictable speech recognition.
• Use 9:16 for vertical short-form content and 1:1 or 4:5 for square or portrait social formats.
• Leave aspect_ratio unset when you want to preserve the original framing.
• Keep the speaker clearly visible when face-following reframing is important.
• Speech that cannot be verified keeps its original audio and is left uncaptioned rather than captioned with uncertain words.

Notes

• videos is required and accepts one source video.
• Maximum input size is 2 GB.
• Typical processing time is approximately 0.7× the source duration, such as about 7 minutes for a 10-minute recording.
• Filler words, false starts, repeated takes, and long pauses may be removed around sentence boundaries.
• If the source already contains burned-in subtitles, generated captions are positioned above them.
• Output aspect ratio follows the source when aspect_ratio is not specified.

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "videos": [
    "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4"
  ],
  "aspect_ratio": "16:9",
  "caption_style": "highlight",
  "language": "auto"
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/ai-video-editor/talking-head" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}")

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id
' >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction finishes.
while true; do
  RESPONSE=$(curl --silent --show-error --fail-with-body \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}")
  RESULT=$(printf '%s' "${RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout|deleted) printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    *) sleep 2 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Response Parameters

Result Request Parameters

Result Response Parameters