Skip to main content

Alchemist vs Tdarr

This is a direct comparison for people who already know what Tdarr is and are evaluating alternatives. It covers the questions that actually come up when switching: licensing, deployment shape, hardware coverage, operational overhead, and where each tool is a better fit.

For current Tdarr behavior, refer to the Tdarr docs. This page only compares against what Tdarr documents publicly, including its server/node/worker model and worker types.

At a glance​

AlchemistTdarr
LicenseGPLv3 (fully open source)See Tdarr's own licensing page
DeploymentSingle binary (also a single Docker container)Central server + node process(es) + workers
Config modelDeclarative — TOML file and UI settingsPlugin stack / flow editor
AV1 targetFirst-class in the planner, uses AV1-capable GPUs when availableSupported via plugins/flows
Hardware accelerationNVENC, Intel Quick Sync, VAAPI, AMD AMF, Apple VideoToolbox, CPU fallbackNVENC, Quick Sync, VAAPI, VideoToolbox (see Tdarr docs)
PlatformsLinux, macOS, Windows, DockerLinux, macOS, Windows, Docker
Non-destructive by defaultYes — delete_source is off by default, output written alongside or to a mirrored rootConfigurable per workflow
Scaling modelScales with a single host's concurrencyScales horizontally with additional nodes

Choose Tdarr if​

  • You want to distribute transcoding across multiple physical machines (the node architecture is Tdarr's core strength).
  • You rely on specific community Tdarr plugins or flows and don't want to re-express them as Alchemist settings.
  • You already have a working Tdarr deployment and no concrete reason to change.

Choose Alchemist if​

  • You want a single binary to deploy. Alchemist is one service. There's no node to register, no server/worker topology to keep healthy, and no plugin stack to debug before a file can be processed.
  • Licensing matters to you. Alchemist is GPLv3 end-to-end — source, binary, and everything it does. No paid tier, no license key, no phone-home.
  • You prefer declarative config over flow editors. The planner decides per file whether to skip, remux, or transcode, using thresholds you set in a TOML file or the UI. See Planner and Skip Decisions.
  • You care about reversibility. Originals are not deleted unless you turn that on explicitly. VMAF can optionally gate the promote step.
  • You're transcoding primarily for Jellyfin or Plex. See Alchemist for Jellyfin.

Practical differences​

Deployment shape​

Tdarr documents a central server, nodes, and workers. That maps well onto a homelab that already spans multiple machines, and poorly onto a single-box setup. If all your media work happens on one host, the server/node/worker split is overhead you carry for a scaling option you may never use.

Alchemist is one process. Concurrency is bounded by the host it runs on. If you want to scale across machines you'd run Alchemist on each machine with its own library roots, or share the library over a network filesystem.

Skip decisions​

Alchemist skips files that are already efficient and surfaces the exact reason in the Skipped tab — BPP below threshold, already in target codec, below minimum file size, or predicted savings below the configured threshold. This is the single most common source of confusion when moving from a flow-based tool; most "why didn't this transcode?" questions resolve against Skip Decisions.

Hardware selection​

Tdarr's docs split CPU and GPU worker behavior and route work based on FFmpeg/HandBrake arguments. Alchemist keeps that decision inside one runtime: it runs a short test encode per backend, caches a valid detection result across repeat boots, and selects one active device using a deterministic scoring policy. The probe log is visible in Settings → Hardware and records exactly why a backend failed.

Open source line​

Alchemist's GPLv3 position is intentionally simple: the public repository is the product. There is no private feature tree, no paid unlock, and no license key. If that matters to you, Alchemist is the cleaner bet. If horizontal worker distribution matters more, Tdarr's architecture is still the thing it does best.

Moving off Tdarr​

There isn't a one-click import — the two tools have different abstractions. The practical path:

  1. Install Alchemist alongside Tdarr (Docker works well).
  2. Point Alchemist at the same library roots.
  3. Run it in dry mode (alchemist plan /path) to see the skip/remux/transcode decision for every file without enqueueing anything. See Installation.
  4. Tune Profiles and Stream Rules until the decisions match what you want.
  5. Disable Tdarr on the same library once you're satisfied.

FAQ​

Is Alchemist a drop-in replacement for Tdarr? No. Tdarr's flow/plugin model doesn't have a one-to-one equivalent in Alchemist. Most real-world Tdarr setups are expressible as Alchemist profiles + stream rules + a target codec, but migration is a config exercise, not an import. See Migrating from Tdarr.

Does Alchemist support distributed nodes? Not today. Alchemist runs as a single process that scales with the host it's on. If you need horizontal scaling across multiple physical machines, that remains a Tdarr strength.

Is Alchemist free? Yes. GPLv3, no paid tier, no license key, no phone-home check. The binary you install is built from the same code in the repository. See Open Source.

Can I run both at the same time during migration? Yes. Running Alchemist in dry-run mode (alchemist plan /path) against the same library roots doesn't touch files — it reports the decision per file. Once satisfied, point Alchemist at the library and disable Tdarr on the same scope. See Installation for the plan subcommand.

Which tool handles AV1 better? Both can target AV1. Alchemist treats AV1 as a first-class output codec in its planner and will pick an AV1-capable hardware encoder when one is present — av1_nvenc (RTX 40 / Ada), av1_qsv (Intel Arc / Meteor Lake+), or av1_vaapi / av1_amf (AMD RDNA 3+). macOS has no AV1 hardware encoder, so AV1 on a Mac uses the CPU (libsvtav1). See AV1.

Does Alchemist have a flow editor? No. Configuration is declarative — a TOML file plus per-library profiles and stream rules. That's the intentional difference; if you specifically want a flow editor, Tdarr or FileFlows are a better fit.

See also​