Video enhancement API for legal & eDiscovery

Why consumer AI tools don't belong in an evidence workflow

Most cloud video-enhancement tools were built for creators, not for custodians of evidence — and two gaps make them a poor fit for legal work.

1. Training clauses buried in the terms of service. Many consumer-grade tools reserve the right to use uploaded content to improve their own AI models unless you find and opt out of a clause in the fine print. For privileged or evidentiary material, uploading footage a vendor may train on is not a defensible position.

2. Vague or indefinite retention. If a vendor cannot tell you exactly when a file is deleted, you cannot represent that it was. 'We probably delete it eventually' does not survive a discovery dispute.

This is no longer only an internal concern. On February 23, 2026, 61 data protection authorities published a joint statement on AI-generated imagery, warning about systems that can generate or manipulate realistic video of identifiable people without consent and calling for clear data-handling transparency and accessible deletion. When your footage shows real witnesses, claimants, or defendants, that is a direct prompt to scrutinize how any vendor trains on, stores, and deletes what you send it.

What legal-grade video enhancement actually requires

Beyond output quality, legal and eDiscovery workflows carry requirements a consumer app rarely meets:

• Provable deletion. A signed deletion certificate and an audit trail, so retention is a record, not a promise.
• Retention you control. A 30-day auto-delete by default that you can shorten, or trigger instantly per file with one API call — aligned to a legal hold or a matter close-out.
• Large, multi-hour files. Full-length depositions and body-cam exports are big; submit-by-URL and pre-signed direct uploads move them without brittle browser uploads.
• Predictable, per-job pricing. Per-second billing and credits that never expire, so a slow month doesn't forfeit budget.
• A DPA on request for your vendor-management file.

These are not niceties. In eDiscovery, an enhanced exhibit still has to be accounted for: where the source went, who could access it, and when it was destroyed. A tool that keeps footage on an unknown retention schedule turns a routine clarity step into a data-handling question you have to answer later. Per-second pricing with credits that never expire also maps cleanly to matter-based accounting — you spend against a specific matter, and unused budget doesn't evaporate at the end of a billing cycle.

Where BetterVideo fits — and where desktop tools can't

Topaz Video AI produces excellent results, but it is a desktop application: it runs on one workstation at a time, on a local GPU, under a per-seat license, and it cannot be called from a review platform or case-management system. If you are building or operating software, you need enhancement as an API, not an app a person runs by hand.

Broad media platforms (for example, Cloudinary) treat upscaling as one feature among many and make no headline privacy guarantee for sensitive footage. Consumer apps in the VanceAI, Media.io, or HitPaw category are accessible and inexpensive, but ship consumer-grade terms — the training-clause and vague-retention problems above.

BetterVideo is the privacy-first, API-first option: one POST /v1/jobs enhances video on GPU (Real-ESRGAN + GFPGAN), your footage is never trained on, never sold, and auto-deleted, and the Secure tier returns a signed deletion certificate with a full audit log. Pricing is per second and credits never expire. See pricing and the API docs, review our approach to data privacy across regulated industries, or — if you also handle claims footage — the insurance claims API. For committed volume, a DPA, and custom retention, talk to sales.

One API call — curl, Python & Node

Submit a video by URL, then poll GET /v1/jobs/{id} until done (or receive a signed webhook), and fetch a time-limited download link:

curl -X POST https://api.bettervideo.io/v1/jobs \
  -H "Authorization: Bearer YOUR_KEY" -H "Content-Type: application/json" \
  -d '{"video_url":"https://.../clip.mp4","resolution":"1080p"}'

import requests
r = requests.post("https://api.bettervideo.io/v1/jobs",
    headers={"Authorization": "Bearer YOUR_KEY"},
    json={"video_url": "https://.../clip.mp4", "resolution": "1080p"})
job = r.json()  # poll GET /v1/jobs/{job['id']} until status == "done"

const res = await fetch("https://api.bettervideo.io/v1/jobs", {
  method: "POST",
  headers: { "Authorization": "Bearer YOUR_KEY", "Content-Type": "application/json" },
  body: JSON.stringify({ video_url: "https://.../clip.mp4", resolution: "1080p" })
});
const job = await res.json();

Privacy is the default, not an upgrade

• Never trained on. Pre-trained, fixed-weight models — your frames are enhanced and returned, never used to train AI.
• Never sold or shared. No third parties, no data brokers, no "anonymized" resale.
• Auto-deleted. Files self-delete (default 30 days), or delete any job instantly with one call.
• Proof on the Secure tier. Every deletion can return a signed deletion certificate, with a full audit log.