Required image plus either voice_script or audio. The schema required list contains image only; speech is a conditional requirement in the guide and property description.Source for Reference guidance (opens in a new tab)
voice defaults to Zephyr (Female); voice_language to English (US). voice_script defaults empty. voice_prompt defaults to "Say the following." and supplies style instructions, not spoken words.Source for Voice controls (opens in a new tab)
Product overview calls the field negative_prompt_strength and gives default 0. Native guide and schema instead use strength_negative_prompt with default 0.5 and range 0–4.Source for Documentation discrepancy (opens in a new tab)
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 15 Oct 2026
Category
Usage
Rate
Unit
Video generation
720p avatar1
$0.025
output video second
Video generation
1080p avatar2
$0.045
output video second
Rate conditions
Per output-video second. Script length, generated speech and uploaded audio may affect final duration; no minimum/rounding rule established.
Per output-video second; no extra voice-specific tariff or standalone audio price established.
Pruna terms retain customer rights in input/output and permit processing to provide the service, plus operational telemetry. Upstream model terms may apply. A universal retention period and zero-training guarantee are not established.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Pruna API
p-video-avatar
Not verified
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
How it works
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
Minimum reusable input
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
Lifetime & refresh
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
How to enable it
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
Usage fields
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
Billing rules
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
Limits & data boundaries
Not collected
No verified prompt-cache mechanism or cached-inference tariff. Asset uploads, conditioning and asynchronous predictions do not establish prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 15 Oct 2026
Voice choices
30 voices
Choose a voice within P-Video-Avatar. These are API voice identifiers, not additional model releases. Source links document each voice.
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Pruna names this hosted video workflow and links the LTX license family. Exact upstream revision and any private post-training are not established. Included explicitly as a managed component; public LTX weights and Pruna optimization-library code are not counted as new private base models.
