Model
Trends
.ai
Model
Trends
.ai
les meilleurs modèles d'IA open source

BitVector: Vector for Discord, Spyglass on the web and Prism chat

Sujet : Applications
Par Captain
Publié le 2026-10-03
How to run hosted models and ComfyUI workflows from Discord with the Vector bot, from any browser with Spyglass, and by chatting with Prism: account linking, /ai render and /ai workflow, initial media for image-to-image and image-to-video, skills, and the fixes for the errors people hit first.
Cette page n'est pas encore traduite ; elle est affichée en anglais.
Informations à jour
Ce guide est un instantané. La source ci-dessous est tenue à jour par ses auteurs ; consultez-la pour les derniers détails.
BitVector documentation (wiki.bitvector.app)

Aperçu

BitVector is a set of three front ends for the same hosted model cloud that
PirateDiffusion
uses. Vector is a Discord bot: slash commands in a server or a direct message come back as images and videos in the channel. Spyglass is a web app for any browser or phone, with a render studio, a model and workflow browser, a gallery and a remix system. Prism is the newest piece, a chat where you describe what you want and Vector proposes the settings and prompt, you approve, and it renders. All three share one account, so a render made in Discord shows up in the Spyglass gallery.
BitVector uses the word Models for what PirateDiffusion calls concepts, and separates them from Workflows. Models are
SD 1.5
and SDXL-family checkpoints (including
Pony
and Illustrious) driven by keyword prompts; Workflows are hosted
ComfyUI
graphs that serve the newer architectures (
Flux
,
Qwen
,
Wan
,
LTX
video, music) with natural-language prompts. Both are called Renderers, and every model page on this site that carries a Run box names the renderer to pick.
This guide condenses the official BitVector wiki into the parts you need to run a model from this catalog: linking the account, the two render commands, feeding an image or video into a workflow, using skills, and the common errors. The wiki is complete, written with care and kept current by the BitVector team; use the link above for anything this page leaves out.

Référence

Nom
Type
Rôle
/graydient link, /graydient confirm-link
Discord
Links a paid hosting account to your BitVector account with a one-time e-mail code. The e-mail is the one used for the hosting plan, not the Discord address, and it is discarded after linking.
/discord (in the PirateDiffusion Telegram bot)
Telegram
The fastest link: the Telegram bot returns a private Linker invitation. Never share that link; linking is semi-permanent.
/ai render
Discord
Renders with a
checkpoint
Model and keyword prompt, the equivalent of PirateDiffusion's /render.
/ai workflow
Discord
Runs a hosted Workflow by name with a prompt, the equivalent of /workflow /run:name. Optional image1 and image2 parameters take direct image links or a recent render.
Apps > Vector > Set Initial Input
Discord
Right-click (long-press) a message with an image or video to make it the starting point of the next workflow, the equivalent of replying to an image in Telegram. Ping @Vector in the message outside a DM.
/skill run, /skill show
Discord
Runs a skill (auto-chosen or named in the skill parameter) and shows a skill's description and template.
/utilities share-render
Discord
Creates a public preview link for a render.
Render page, Browse page, Gallery
Spyglass
The render studio with every workflow setting, the searchable list of models, workflows and skills, and your results with a Remix button.
Render Mode: Manual or Skill
Spyglass
Manual picks a renderer and exposes its settings; Skill takes a description, proposes a command, and runs it on Approve & Run.
Input Media: Target Image/Video URL, image1, image2
Spyglass
Appear automatically when the chosen workflow needs them. Paste a direct link, pick a recent render or upload a file (kept 60 days).
Render Approvals
Prism
Prism always shows the proposed settings and prompt before anything is submitted; you approve each render.
Adult metadata toggle
Prism
Vector's catalog search hides models and workflows flagged NSFW unless this account-bound toggle is on.
Blueprints manager
Prism
Prism-native skills. Create one yourself in the Blueprints manager, or accept one Vector proposes after you approve it. Blueprints are tuned for Vector's tools and work only in Prism; Discord and Spyglass keep using the PirateDiffusion skills of the shared account.
AD
Aucune installation requise – exécutez les modèles IA dans le cloud
BitVector Prism est la façon la plus simple de commencer : choisissez un modèle, tapez un prompt et générez dans une application web épurée, sans rien configurer. BitVector est aussi disponible sur Discord et sur le web (SpyGlass).

Pas à pas

  1. Get the requirements: a paid hosting plan, and a Discord or Telegram account. Spyglass needs only a browser.
  2. Link the account. From Telegram, send /discord to the PirateDiffusion bot and open the Linker link. From Discord, run /graydient link, enter the plan e-mail, then /graydient confirm-link with the six-digit code. In Spyglass, follow the red banner.
  3. Make a first image in Discord: /ai workflow, workflow zimage-turbo, then a descriptive prompt. The result embed carries the exact command for re-running and editing.
  4. Make the same image in Spyglass: Render, type the prompt, Select a Model or Workflow, Workflows tab, search zimage, Select, scroll down, Generate Render.
  5. Run a model from this catalog: read the renderer name in the model's Run box; use /ai render for checkpoints and /ai workflow for workflow-served models.
  6. Feed in an image: in Discord, send the image with @Vector pinged (not needed in a DM), then Apps > Vector > Set Initial Input and run the workflow. In Spyglass, select the workflow and use the picture or upload buttons under Input Media.
  7. Try skills: /skill run with just a prompt in Discord, or Render Mode: Skill in Spyglass, review the proposed command, Approve & Run, then vote.
  8. Chat in Prism when you do not know which model to use: describe the result, read the proposal, approve, iterate. When Vector offers to turn a routine you repeat into a Blueprint, approve it, or make one yourself in the Blueprints manager.

Exemples

First render in Discord

/ai workflow workflow: zimage-turbo prompt: A cinematic picture of a dark living room, a wooden coffee table in front of a TV, an open pizza box, neon light strips glowing on the wall behind the TV

A checkpoint model with keyword tags

/ai render model: realvis4-xl prompt: portrait of a fisherman mending nets at dawn, harbour fog, high quality, masterpiece negative: blurry, low resolution, deformed

Image to video

Send the image to @Vector, right-click > Apps > Vector > Set Initial Input
/ai workflow workflow: animate prompt: the camera slowly pushes in while the waves roll

Skills

/skill run prompt: a warrior standing in the rain, cinematic
/skill show skill: chroma1-flash
In Prism: "Make a Blueprint that upscales my last render and adds film grain" and approve the proposal

Astuces

  • Discord posts results where you ran the command. Use a DM with Vector or an age-restricted channel for anything that could be NSFW; that is Discord's rule, not only BitVector's.
  • Workflows have a three-minute compute limit. Huge sizes or long videos time out; lowering resolution or frames is the fix and there is no penalty for trying.
  • Workflows ship with sensible defaults. Change one setting at a time in Spyglass; you never have to touch them all.
  • For two- or three-image workflows (swap, blend), Spyglass is much easier than Discord: the image1 and image2 fields appear with upload buttons.
  • Read the workflow description for image order. Replying in Telegram equals Set Initial Input; /image1:name equals the image1 parameter. Mixing them reverses results.
  • Install Spyglass as a progressive web app on a phone; the system Share menu then sends images straight into your uploads.
  • Skills remember your last render: "upscale this video" or "edit this last image to..." works without choosing anything.
  • Renders made through skills can still be remixed from the gallery; some settings translate lossily, so re-select initial media after remixing.
  • Two kinds of skills: PirateDiffusion skills (made on the web or in Telegram) work in Discord and Spyglass; Blueprints are made in Prism and work only there.

Dépannage

Render fails right after linking or before linking

Pourquoi cela arrive
No paid hosting account is linked to the BitVector account.

Comment le corriger
Run /graydient link in Discord or follow the red banner in Spyglass. The plan name appears under your profile picture when linking has worked.

Vector says no media was detected

Pourquoi cela arrive
@Vector was not pinged in the message carrying the image, so Discord hid the content from the bot; or the link is not a direct media link.

Comment le corriger
Re-send the image with @Vector in the same message (not needed in a DM), then Set Initial Input. For links, use Copy Link on the image, not Copy Message Link.

You don't have an initial context set

Pourquoi cela arrive
The message had several attachments and none was chosen, or the workflow needs image1/image2 instead of a reply.

Comment le corriger
Pick one attachment in the dropdown Vector shows, or pass direct links in the image1 and image2 parameters as the workflow description says.

Timeout error on a workflow

Pourquoi cela arrive
The combination of resolution, frames and steps exceeded the three-minute limit.

Comment le corriger
Lower the settings and run again; defaults are tuned to finish in time.

Spyglass shows a generic failure after a workflow run

Pourquoi cela arrive
The workflow needed initial media and none was provided; Spyglass does not yet block this.

Comment le corriger
Select the workflow, fill the Target Image/Video URL under Input Media with a recent render or upload, and run again.

A model or skill cannot be found by Vector in Prism

Pourquoi cela arrive
It is flagged NSFW and the adult metadata filter is on.

Comment le corriger
Settings at the bottom left of the Prism sidebar, toggle Adult metadata (account-bound).

Model list in Spyglass is empty or stale

Pourquoi cela arrive
The catalog cache has not synced yet.

Comment le corriger
Wait for the syncing dialog to finish, or press Refresh at the top right (allowed every five minutes).

Strange field named init_image_url

Pourquoi cela arrive
An API-only field leaked into the Spyglass settings list.

Comment le corriger
Ignore it; the fields you set have the recent-render and upload buttons next to them. slotX fields are the only unusual ones meant for you.

AD
PirateDiffusion
Aucune installation requise – exécutez les modèles IA dans le cloud
PirateDiffusion est réservé à Telegram et conçu pour les pros : des milliers de modèles, LoRA et workflows pilotés par commandes de chat, avec génération illimitée à prix fixe.

Questions

Do I need PirateDiffusion to use BitVector?

No, but they share the hosting account. Linking is quickest through the PirateDiffusion Telegram bot's /discord command, and a one-time code by e-mail works without it.

Which models can I run?

Most of the hosted catalog: SD 1.5 and SDXL-family checkpoints and
LoRAs
as Models, and Flux, Qwen, Wan, LTX and audio models as Workflows. A few exotic workflows are not exposed; the Spyglass Browse page is the authoritative list.

What is Prism compared to Spyglass?

Spyglass is a studio where you choose the renderer and settings. Prism is a chat where Vector chooses them, shows you the proposal and renders on your approval. Prism is experimental and changes often.

What is Beacon?

An earlier BitVector tool that exposes rendering to your own LLM agent through an MCP server. Prism is the batteries-included version of the same idea.

Is NSFW allowed?

Within both platforms' terms, yes, but Discord requires age-restricted channels or a DM, and Prism hides NSFW-flagged models from Vector unless the adult metadata toggle is on.

Can I create skills in BitVector?

Yes, in Prism. Prism has its own skills called Blueprints: you create one in the Blueprints manager, or Vector proposes one while you chat and builds it when you approve. They are tuned for Vector's tools and work only in Prism. Vector for Discord and Spyglass do not run Blueprints; they use the PirateDiffusion skills that come with the shared hosting account.

Liens et sources

Modèles de ce guide

Écrit par
Captain

Guides associés

Services cloud
PirateDiffusion: run any model from Telegram, no GPU needed
The commands that matter in the PirateDiffusion Telegram bot: /render with model trigger words, (( )) and [[ ]] weighting, #recipes, /adetailer, /highdef and /facelift upscaling, /remix, /inpaint, ComfyUI workflows with /wf /run:, and the new // skills that pick the model for you.
Par Captain
2026-10-03
ComfyUI
ComfyUI nodes: documentation and troubleshooting
Plain-language documentation for the ComfyUI nodes people use most: what each node does, every input and output, how to wire it, the settings that matter, and the errors it throws with the fix for each one.
Par Captain
2026-06-02
Erreurs et solutions
ComfyUI errors and how to fix them: red nodes, missing models, CUDA out of memory
The ComfyUI error messages people hit first, what each one means and the fix that works: missing custom nodes (red boxes), "Prompt outputs failed validation", CUDA out of memory, wrong model type in a loader, mat1 and mat2 shape errors, header deserialization, torch and xformers mismatches.
Par Captain
2026-05-12
Modèles
Flux vs SDXL vs SD 1.5 vs Qwen vs Z-Image: which model family to use in 2026
A practical comparison of the open image model families: prompt following, realism, anime and illustration, text rendering, speed, VRAM, LoRA ecosystem and licence, with a recommendation for each kind of work and hardware.
Par Captain
2026-05-22

Autres guides

Modèles
FLUX.1 Dev: how to prompt it, best settings and creative ideas
A practical guide to FLUX.1 Dev by Black Forest Labs: natural-language prompting, guidance and step settings, LoRA stacking, text rendering and prompt ideas that play to its strengths.
Par Captain
2026-09-30
Modèles
Stable Diffusion XL 1.0: prompts, settings and what it still does best
How to get the most out of the official SDXL 1.0 base model: native resolutions, CFG and sampler settings, the refiner, negative prompts and ideas for styles where SDXL still shines.
Par Captain
2026-10-01
Modèles
Stable Diffusion 1.5: the classic model, prompted properly
Stable Diffusion 1.5 is still worth knowing: the right resolution, CFG and sampler, how to use its huge library of LoRAs and embeddings, and creative prompt ideas suited to a 512-pixel model.
Par Captain
2026-10-02
Modèles
FLUX.2 Dev: prompting the newest Black Forest Labs model
FLUX.2 Dev brings larger prompts, better text, stronger realism and multi-reference editing. This guide covers the turbo workflow, guidance and resolution settings, prompt structure and ideas that use its new strengths.
Par Captain
2026-10-03
Modèles
Qwen-Image: long prompts, perfect text and bilingual posters
Qwen-Image is the model to reach for when the words in the picture matter. Learn how to write its long descriptive prompts, render English and Chinese text accurately, set CFG and steps, and explore layout-heavy creative ideas.
Par Captain
2026-09-29
Modèles
SDXL-Lightning: four-step generation on any SDXL checkpoint
ByteDance's SDXL-Lightning LoRA turns a 30-step SDXL render into a 4-step one. Learn the step counts, the CFG you must use, the sampler, and how to combine it with your favorite checkpoints and style LoRAs.
Par Captain
2026-09-30