--- license: unknown pipeline_tag: text-to-image tags: - text-to-image - image-to-image - image-editing - multi-reference - distilled - turbo --- # Mage-Flow Turbo + Edit Turbo 4-step photographic text-to-image (**≈3 s** per 1024² image on an RTX 5090) and a companion instruction-editing model that takes up to **three** reference images. > **What this repo is:** the Mage-Flow turbo and turbo-edit transformers, a > shared Qwen2.5-VL text encoder and VAE — weights only, not a retrain. The > settings below are the values these weights are actually run with day to > day. --- ## Samples — `mage-flow-turbo` (txt2img) Four steps, guidance 1.0. Photorealism is this model's strength.
prompt: a red apple on a rustic wooden table beside a window, soft daylight, fine skin texture and wood grain, studio photograph, sharp focus — 1024×1024, 4 steps, guidance 1.0, seed 12345prompt: portrait of an older fisherman with a weathered face, natural window light, shallow depth of field, photorealistic — 1024×1024, 4 steps, guidance 1.0, seed 99
**prompt:** `a red apple on a rustic wooden table beside a window, soft daylight, fine skin texture and wood grain, studio photograph, sharp focus` — 1024×1024, 4 steps, guidance 1.0, seed 12345**prompt:** `portrait of an older fisherman with a weathered face, natural window light, shallow depth of field, photorealistic` — 1024×1024, 4 steps, guidance 1.0, seed 99
prompt: a rain-slicked Tokyo street at night, neon signs reflecting in puddles, cinematic, 35mm photograph — 1024×1024, 4 steps, guidance 1.0, seed 1234
**prompt:** `a rain-slicked Tokyo street at night, neon signs reflecting in puddles, cinematic, 35mm photograph` — 1024×1024, 4 steps, guidance 1.0, seed 1234
--- ## Editing — `mage-flow-edit-turbo` (img2img) Plain-language instructions against a reference image. Identity, pose, lighting and camera are preserved; only what the instruction names changes.
referenceedit: put a brown flat cap on his headedit: change the background to a sunny beach
**reference** — the seed-99 portrait above**instruction:** `put a brown flat cap on his head` — 4 steps, seed 7 (≈88 s)**instruction:** `change the background to a sunny beach with the ocean behind him` — 4 steps, seed 7 (≈60 s)
Both edits keep the subject essentially pixel-identical — same skin detail, same jacket, same catchlights — while adding an object in one case and replacing the entire environment in the other. Up to **three** reference images may be supplied together (subject, style, context). ## Recommended settings Values these builds are run with in practice. | Parameter | `mage-flow-turbo` | `mage-flow-edit-turbo` | Meaning | |---|---|---|---| | `width` × `height` | `1024` × `1024` | follows the reference | Output size | | `steps` | `4` | `4` | Denoising steps | | `guidance` | `1.0` | `1.0` | Distilled — CFG is not used | | `max_size` | — | `1024` | Longest edge the reference is fitted to | | `vl_condition_long_edge` | — | `384` | Resolution the VL encoder sees the reference at | **Supported modes:** `txt2img` (turbo), `img2img` / instruction editing (edit-turbo, 1–3 references) ### Notes and gotchas - **Four steps is correct.** Both variants are step-distilled at guidance 1.0; raising either is not the lever for quality. - **Editing costs far more than generating** — ≈60–90 s versus ≈3 s, because the reference must be encoded through the VL tower before denoising begins. That asymmetry is inherent to the architecture, not a misconfiguration. - **Write edits as instructions, not descriptions** — `put a brown flat cap on his head`, not `a man wearing a brown flat cap`. - **`torchvision` is a hard dependency** of the Mage-Flow pipeline (the VL image preprocessing path) even though the rest of this stack does not need it. Install the build matching your torch (`torchvision==0.28.0` for torch 2.13.0+cu130) or loading fails with `ModuleNotFoundError`. - **Three transformers ship here**: `base` (30 steps, guidance 5.0 — the undistilled model), `turbo` and `turbo-edit` (4 steps, guidance 1.0). The samples on this card are from the two turbo variants; the base transformer is included for anyone who wants the slower, higher-guidance path. - Each transformer directory is self-contained — download only the variant you intend to run (8.2 GB each) plus the shared encoder, VAE and scheduler. --- ## Layout Components ship as separate directories under `split/`: one transformer directory per variant (`turbo`, `turbo-edit`), plus the shared text encoder, VAE and scheduler config. Loaders that accept a diffusers-style component tree can consume this directly. --- ## Files | Path | Size | Role | |---|---|---| | `split/transformer/turbo/diffusion_pytorch_model.safetensors` | 8.23 GB | txt2img transformer (4-step distilled) — **the tier these settings were measured on** | | `split/transformer/turbo-edit/diffusion_pytorch_model.safetensors` | 8.23 GB | instruction-editing transformer (4-step distilled) | | `split/transformer/base/diffusion_pytorch_model.safetensors` | 8.23 GB | undistilled transformer (30 steps, guidance 5.0) | | `split/text_encoder/` | 8.88 GB | Qwen2.5-VL text/vision encoder (2 shards) | | `split/vae/diffusion_pytorch_model.safetensors` | 345 MB | VAE | | `split/scheduler/scheduler_config.json` | 169 B | scheduler config | --- ## Provenance - **Upstream base model:** Mage-Flow (turbo and turbo-edit distillations) - **This build:** redistributed as a split component tree with the shared encoder and VAE alongside. Weights are not retrained here. - **License:** left as `unknown` in this repo's metadata. Refer to the upstream model's license for redistribution and commercial-use terms.