Agency Guide

Preset selection, team workflow, and volume scaling.

Written for agency operators managing multiple accounts and distributing the same content across platforms at volume. All processing happens locally — your content never leaves your machine.

v1.2.05 agency presetsAgency plan

Which Preset for Which Platform

Duplyfi ships with five presets designed for agency multi-platform workflows. Each is tuned to the uniquification depth that a given platform's content pipeline actually requires — not more, not less. Over-processing wastes time; under-processing leaves content variation on the table.

Standard

Daily Driver

Platform: All Platforms  ·  Copies: 5

5 copies

Your default preset for multi-platform distribution. Use this when posting the same source clip to TikTok, Instagram, and X simultaneously.

Effects enabled
  • Zoom (3–6%): Unique spatial framing per copy via randomized crop factor
  • Cut Start (3–6%) + Cut End (3–7%): Changes temporal boundaries for perceptual differentiation
  • Noise (level 4–8, unique seed per file): Different grain pattern on every output
  • Pitch Shift (±1 semitone): Audio pitch shift with speed compensation
  • Spectral Tilt: Per-file EQ adjustments around 8kHz and 150Hz — adjusts audio frequency profile
  • Luma Curves: Randomized per-file tone curve, invisible to viewers
  • US Metadata: Randomized device metadata (model, GPS, iOS/Android version, timestamp)

When to use: Use when distributing to multiple platforms from the same clip in the same posting cycle. For TikTok accounts where you want maximum uniquification depth, switch to TikTok Heavy for those files.

TikTok Heavy

Maximum Depth

Platform: TikTok  ·  Copies: 3

3 copies

Everything in Standard, plus deeper temporal trimming, Perspective Warp, Blurred Border, and Reverb. Use for high-value TikTok accounts or when distributing the same clip across a large number of TikTok accounts.

Effects enabled
  • All Standard effects, plus:
  • Larger cuts (5–10% start, 5–12% end): More aggressive temporal trimming
  • Perspective Warp: Sub-visual projective transform, 0–1.5% corner shift per file
  • Blurred Border (20–30px): Adds new visual content at frame edges
  • Reverb (60–100ms, moderate decay): Modifies audio frequency response timing
  • Higher noise levels (5–10)

When to use: Use for TikTok accounts where maximum content variation depth is required. For routine multi-platform batches, Standard is sufficient.

Instagram

Visual Focus

Platform: Instagram Reels  ·  Copies: 3

3 copies

Instagram's content analysis (Meta TMK+PDQF) is sensitive to crop and color changes but places less emphasis on audio fingerprinting than TikTok does. This preset prioritizes visual uniquification to match that profile.

Effects enabled
  • Brightness (0.03–0.07): Per-file luminance shift
  • Saturation (1.02–1.08): Slight per-file color boost
  • Vignette (0.2–0.4): Radial darkening that adds new visual content at frame edges
  • Noise (level 2–5, unique seed per file): Lighter than TikTok presets
  • Cut Start (2–5%): Small temporal trim
  • Luma Curves: Per-file randomized tone curve
  • US Metadata: Device metadata injection

When to use: Use when posting only to Instagram to minimize processing time. For multi-platform batches (Instagram + TikTok + X), use Standard — it covers Instagram and includes additional audio variation that TikTok requires.

Reddit Safe

Light

Platform: Reddit  ·  Copies: 2

2 copies

Reddit re-encodes all uploaded video regardless of source. Light uniquification handles the file-level check before re-encoding. For Reddit, operational hygiene (account isolation, posting cadence) matters more than processing depth.

Effects enabled
  • Brightness (0.02–0.04): Minimal brightness shift
  • Saturation (1.01–1.04): Minimal saturation shift
  • Noise (level 1–3, unique seed per file): Light noise for file-level variation
  • US Metadata: Device metadata injection

When to use: Use only for Reddit. Applying a heavier preset to Reddit provides no additional benefit because Reddit re-encodes your file. For Reddit, your operational stack (proxies, account isolation) determines distribution performance.

High-Volume Speed

Throughput

Platform: Any (500+ files)  ·  Copies: 5

5 copies

Agencies processing large content libraries need consistent throughput more than maximum per-file variation depth. This preset is optimized for batches of 500+ files.

Effects enabled
  • Noise (level 3–6, unique seed per file): Sufficient for file-level hash differentiation
  • Cut Start (2–4%) + Cut End (2–4%): Minimal temporal trim
  • US Metadata: Device metadata injection

When to use: Use when processing large backlogs quickly for lower-priority platforms. Estimated processing time: 8–12 seconds per file on modern hardware. A 500-file batch on an 8-core machine with 4 parallel workers takes roughly 17–25 minutes.

Want to understand what each effect does and which content dimension it modifies?

Effect reference →

Multi-Account Best Practices

File-level uniquification handles content variation at the file layer. These practices address the operational layer that agencies managing multiple accounts need to consider separately.

How many copies per source file

The number of copies you need is determined by how many unique accounts will receive the same source clip. A typical agency setup: 10 accounts, each posting to TikTok, Instagram, and X on the same day.

10
accounts
3
platforms
30
unique files needed

10 accounts × 3 platforms = 30 unique files per source clip per posting cycle

Do not reuse the same output file across two different accounts on the same platform. Each output file should go to exactly one account. If your accounts are clustered into groups, use platform-specific presets per group.

Staggering uploads

Avoid posting from all accounts simultaneously. Platforms monitor for coordinated posting patterns across accounts. A practical stagger schedule for a 10-account batch:

Window 1Accounts 1–3First 30-minute window
Window 2Accounts 4–72–4 hours later
Window 3Accounts 8–10Following day

The exact timing matters less than consistency and the absence of clustering. Ten accounts posting at 9:03, 9:04, 9:06, 9:08, 9:09 from the same IP block has a behavioral signature that is distinct from ten unrelated operators posting independently.

File naming

Duplyfi outputs files with random UUID-based names (e.g., 3f8a2c1d-...mp4). This is intentional. Do not rename output files to match the original filename or to include account numbers, team identifiers, or date codes.

Platforms read the filename embedded in the upload request. A consistent naming pattern across uploads from different accounts creates a behavioral fingerprint at the upload metadata layer. Random UUID names provide no deducible pattern across accounts.

Rotating presets across accounts

For high-value accounts, rotate between Standard and TikTok Heavy across posting cycles. Variation in the effect profile itself adds another dimension of differentiation between posting events.

Accounts 1–2TikTok HeavyMaximum depth
Accounts 3–4StandardStandard depth
Account 5TikTok HeavyRotate with accounts 1–2 next cycle

Team Setup (Agency Plan)

The Agency plan includes 10 device slots. One slot is consumed per machine logged in to your agency account.

Adding devices

Have each team member install Duplyfi on their machine and log in with the agency account credentials. Login automatically binds their hardware ID to your license. No manual key activation is needed. If you have 10 operators running processing in parallel across different account batches, all 10 machines can work simultaneously on the same license.

Cloud preset sync

Agency plan includes cloud preset sync. Build your preset library once — customized presets for specific content niches, platform-specific configurations, high-volume speed settings — and every device on your agency account automatically has access to the same preset set. Designate one senior operator as the preset owner; all other devices sync automatically.

Managing device slots

If a team member leaves or a machine is retired, free up the slot from the web dashboard (Dashboard > Devices > Deactivate). The freed slot is immediately available for a new device. If you temporarily exceed 10 devices (e.g., onboarding a new operator before offboarding a departing one), login on the 11th device will be blocked. Deactivate one existing device first.

Credentials policy

Do not share login credentials across two operators working simultaneously — each requires their own active device slot. Account credentials control license enforcement; two active sessions from the same credentials count as two device slots.

Volume Processing Tips

Parallel processing, hardware selection, and workflow structure for agencies processing hundreds of files per week.

Parallel processing

Duplyfi processes multiple files concurrently. All plans support up to 4 parallel workers, auto-scaled based on your CPU cores and available RAM. Systems with less than 12 GB RAM are capped at 2 workers automatically. There is no tier gating on worker count.

CPU

Primary throughput lever. Each processing worker uses roughly one physical core. 4 cores = 4 workers at near-100% utilization. 8 cores leaves capacity for other applications.

RAM

Primary constraint at high concurrency. Each worker requires 200–400MB peak for 1080p source. 4 workers parallel = 800MB–1.6GB peak. 12 GB RAM or more: up to 4 workers. Less than 12 GB RAM: capped at 2 workers automatically.

GPU

Auto-detected at startup. NVIDIA NVENC (Win/Linux): roughly doubles throughput per worker. Apple Silicon VideoToolbox (macOS). Intel/AMD VAAPI (Linux). Falls back to CPU automatically if no GPU detected.

500+ files per week workflow

10 accounts × 3 platforms × 20 source clips per week = 600 output files. Use this workflow:

  1. 1Organize source files by platform destination before processing. Group all TikTok-bound clips, all Instagram-bound clips, and all X-bound clips into separate input folders.
  2. 2Run one batch per platform group using the appropriate preset (TikTok Heavy for TikTok, Instagram for Instagram, Standard for X and Reddit).
  3. 3Set copies to match the number of accounts receiving each clip on that platform (e.g., 10 copies if all 10 accounts post to TikTok).
  4. 4Route output files to platform-specific output folders. Your upload operator then picks from the appropriate folder for each account.
  5. 5Do not mix output files across platforms. A file processed with the Instagram preset has different effect parameters than one processed with TikTok Heavy.
For batches above 500 files, run in two sessions rather than one. Start the first session before end of business; run the second session the following morning. This keeps individual batch sizes manageable and lets you inspect output quality from the first batch before committing the second.

System Requirements

Minimum specs for running Duplyfi, and recommended specs for high-volume agency processing.

ComponentMinimumRecommended
CPU4-core (Intel i5 8th gen, AMD Ryzen 5 3000, or Apple M1)8-core (Intel i7/i9 10th gen+, AMD Ryzen 7/9 5000+, or Apple M2/M3)
RAM4 GB8 GB+
Disk2 GB free (plus space for outputs)SSD with 10 GB+ free for active batches
OSWindows 10 (64-bit), macOS 12 Monterey, or Ubuntu 20.04+Latest stable release of your operating system

GPU acceleration

GPU acceleration is detected automatically at startup. No configuration is required. If your GPU is available and supported, Duplyfi will use it automatically. If not supported or not detected, processing falls back to CPU encoding with no change to output quality.

NVIDIA (Windows, Linux)
NVENC via CUDA

GeForce GTX 10 series or newer. Supports up to 4 concurrent NVENC sessions (hardware maximum). Operators with NVIDIA cards should expect 8–14s/file vs 18–35s/file CPU-only.

Apple Silicon (macOS)
VideoToolbox

Available on M1, M2, and M3 chips via the onboard Media Engine. Detected automatically on compatible macOS builds.

Intel/AMD (Linux)
VAAPI

Available on most modern Intel processors and AMD APUs running Linux.

Questions about your agency workflow? Reach us on Telegram or email support@duplyfi.com.