Introduction
Upload a portrait and a voice recording. An optional prompt can steer expression and movement; otherwise the delivery follows the audio.
| Field | Value |
|---|---|
| Input | Image, Audio |
| Output | Video |
| Input combination | One portrait and one voice recording |
| Portrait formats | JPEG, PNG or WebP |
| Audio duration limit | Up to 120 seconds |
| Delivery prompt | Optional; 1 to 2,000 characters when supplied |
| Output resolution choices | 480p or 720p |
| Default resolution | 720p |
| Relative resolution rate | For equal audio length: 720p rate = 2 x 480p rate |
| Billing basis | Each started second of the input recording |
| Minimum charge | Four seconds |
| Unfinished runs | No charge |
| Recorded test input | A portrait and 9.2 seconds of audio, with 480p selected |
| Recorded test output | MP4, 556 x 740 pixels, 9.38 seconds |
Key features & highlights
A short recording returned as MP4
Our November 18, 2026 run paired a portrait with a 9.2-second voice clip and returned a 556 x 740 MP4 lasting 9.38 seconds. The request selected 480p and asked for warm speech with calm, natural head movement. Play the example with sound to inspect the delivery.
Direct the speaking style
Use up to 2,000 characters to guide expression and movement. Put the spoken words in the recording and describe the performance in the prompt: welcoming, composed or focused, with the head movement you want to see. Keep the direction concise enough to check against the finished clip.
Choose the rate before submitting
Output choices are 480p and 720p, with 720p selected by default. At an equal input-audio length, 720p costs twice as much as 480p. Billing counts each started second of the input recording and applies a four-second minimum; a run that does not finish carries no charge.
Request parameters
| Parameter | Required | Description |
|---|---|---|
image_url | Yes | The portrait to animate. at least 1 characters |
audio_url | Yes | The voice recording, up to 120 seconds. Recordings under 4 seconds are billed at the 4-second minimum. at least 1 characters |
prompt | No | Optional guidance for expression and movement. at least 1 characters · at most 2000 characters |
resolution | No | 480p or 720p (default). 720p costs twice as much. |
Pricing
Talking Avatar Flash is billed per started second of input audio, with a four-second minimum. Runs that do not finish carry no charge.
| Specification | Price |
|---|---|
| $0.162 / second |
| $0.324 / second |
Use cases
A recorded introduction
Pair a presenter portrait with a prepared welcome recording for a course or event. Give the optional direction a narrow job, such as a friendly expression with restrained head movement, then review the delivery before sharing it.
A voice-led product briefing
Use a professional spokesperson portrait with a recording that explains the product. Settle the spoken wording in the audio before generation, and use the prompt to describe the expression and movement you want to try.
A guide for an exhibition
Pair a guide portrait with a prepared recording about an exhibit. Describe a composed expression and restrained head movement, then review the video alongside the narration before adding it to the exhibition materials.
A workshop welcome
Give a craft instructor portrait a recorded introduction to a lesson. Use the movement prompt to request a relaxed delivery, and check that the finished clip fits the pace of the teaching materials.
Prompt guide & pro tips
Prepare the portrait and recording
Choose a JPEG, PNG or WebP portrait and a voice recording up to 120 seconds long. Listen through the audio and settle the spoken wording before uploading.
Guide the expression and movement
Add a short direction for the delivery, such as a welcoming expression with restrained head movement. The prompt is optional and accepts up to 2,000 characters.
Select the output resolution
Choose 480p or 720p; the default is 720p. For the same input-audio duration, 720p costs twice as much as 480p.
Watch the completed video
Play the clip with sound and compare its mouth movement, expression and ending with your recording. Review the delivery in the context where you plan to use it.
Give a portrait your recorded speech
Talking Avatar Flash animates a still portrait from a voice recording. Supply the image and audio, choose 480p or 720p, and add optional guidance for expression and movement. This workflow suits a course welcome, a presenter introduction or a product briefing whose spoken wording is already prepared.
The recording carries the speech. Treat the prompt as direction for the performance: how the face should look and how the head should move while the person speaks. Keeping those roles clear makes it easier to prepare the inputs and assess the returned video.
Prepare the portrait and audio
Provide a JPEG, PNG or WebP portrait in image_url and a voice recording in audio_url. Both fields require a nonempty value. The audio can be up to 120 seconds long, so check the complete file duration before submitting it.
Choose a portrait where you can clearly inspect the face. Listen to the recording from beginning to end and settle the wording, pauses and ending before generation. Remove an unwanted take or accidental silence while preparing the audio rather than trying to correct the spoken content through the movement prompt.
Input fields
Use the required pair of URLs for the source material, then set the optional fields according to the delivery and output height you want.
| Field | Purpose | Accepted value |
|---|---|---|
image_url | Portrait to animate | Required, nonempty string; JPEG, PNG or WebP |
audio_url | Voice recording that supplies the speech | Required, nonempty string; up to 120 seconds |
prompt | Guidance for expression and movement | Optional; 1 to 2,000 characters when supplied |
resolution | Output height setting | 480p or 720p; default 720p |
Write direction you can review
Describe the expression and movement in a short, coherent instruction. For a welcome, you might ask for a friendly expression and steady head movement. For a briefing, a focused expression with small, measured movements gives you a different direction to assess. Match the instruction to the purpose of the recording.
Avoid filling the prompt with a second version of the script. If a product name, lesson detail or greeting needs to change, revise the audio before submitting the next request. Use the text field for the delivery you want to see while that recording plays.
When you do not need written direction, omit prompt. If you supply it, keep it within the 2,000-character limit and give it a nonempty value. To compare two directions, keep the portrait, recording and resolution unchanged and revise a specific expression or movement instruction. Save both prompts with their results so you can review the difference.
Choose the resolution and plan the charge
The available output height settings are 480p and 720p. Omitting resolution selects 720p. At the same input-audio duration, 720p costs twice as much as 480p. Choose the setting before submission and use the price shown on this page to plan the request.
Billing counts each started second of input audio and applies a four-second minimum. A recording shorter than four seconds still uses that minimum. For longer recordings, a partial second counts as a started second. Base your estimate on the source recording and selected resolution rather than the duration of the returned video. Runs that do not finish carry no charge.
Our portrait-and-audio result
On November 18, 2026, our night-sky-talk run paired a portrait with a 9.2-second voice clip at the 480p setting. It returned a 556 x 740 MP4 lasting 9.38 seconds. The movement instruction was:
He speaks warmly with calm, natural head movement.
Open the example with sound to inspect how the speaking portrait presents the recording. The input audio and returned video have different measured durations in this run; use the input recording when working out the audio-based charge.
Review before sharing
Watch the completed clip alongside the source portrait and recording. Check the mouth movement against the speech, examine the expression during pauses, and listen through the ending. Consider whether the delivery fits its placement in your lesson, presentation or product briefing.
If you want to revise the delivery, identify a particular expression or movement to change. If the speech itself needs correction, prepare a revised recording. Keep the inputs, prompt and resolution with the result so the next request has a clear starting point.
Compare related talking-avatar options
Choose this endpoint when your task starts with a portrait and a prepared recording of up to 120 seconds, and its 480p or 720p settings fit your intended use. For another talking-avatar workflow, compare the pages for talking-avatar/v1 and infinitetalk/v1. Review their input fields, settings and displayed prices against the source material you plan to use before choosing a family.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "spicyapi/talking-avatar-flash/talking-avatar",
"input": {
"audio_url": "https://example.com/reference.mp3",
"image_url": "https://example.com/reference.png",
"resolution": "720p"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.
