# Text, speech, image and music models behind one local executable

> Text, speech, image and music models behind one local executable: 1 companies fully do it, 4 with a limit, 110 carry the tag. Every listing in Local AI with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/local-ai.multimodal-local-studio/
- Markdown: https://www.anchorterminal.com/niches/local-ai.multimodal-local-studio/index.md (~1,450 tokens)
- Slim: https://www.anchorterminal.com/niches/local-ai.multimodal-local-studio/index.min.md (~280 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/local-ai.multimodal-local-studio/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

One company's docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 1 | 4 | 110 | 2 | 3 |

## What counts as done

One executable serves a chat completion, a speech-to-text request and an image generation through documented routes, with the owner choosing which to load.

Measured, for our own tests: each returns within the time the vendor states.

How agents ask for it: try a bit of everything locally; an AI radio station that writes, speaks and plays its own music; generate an image and a voice line from the same server

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **LocalAI (Ettore Di Giacinto and the LocalAI team)**: One server with per-modality backends serves chat, audio-to-text, text-to-audio, image and video generation through OpenAI, Anthropic and ElevenLabs shaped routes, each backend pulled when a model needs it.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Berth (Trevin Tindall (T-Crypt on GitHub))**: Lets llama.cpp, ComfyUI and others take turns on one card with a priority queue, which answers the shared-GPU blocker, but it serves no speech route of its own and ComfyUI's API is proxied, not unified.
- **KoboldCpp (LostRuins (Concedo))**: One executable serves chat, /v1/audio/transcriptions, /v1/audio/speech, /v1/images/generations and music generation, each loaded by flag. --password covers text routes only; the help text says image endpoints are not secured.
- **Lemonade (Advanced Micro Devices, Inc.)**: One server does chat, /v1/audio/transcriptions, text-to-speech and image generation (omni collections and MCP tools). Images arrive through tool calls or MCP rather than a plain images route, and one GPU is shared.
- **TextGen (oobabooga)**: Chat, vision input and /v1/images/generations with a loaded image model are documented; speech routes are not in the API doc.

## Couldn't check

The docs were closed to us, or silent. These carry no mark either way.

- **Docker Model Runner (Docker, Inc.)**: The image.generate tag could not be checked: the README shows no image route and the docs site is paused for automated reading.

- not this job: AnythingLLM (Mintplex Labs, Inc.): Chat and document app; image generation and transcription route to providers rather than being served.
- not this job: BitRun (Abhishek Khanra (IshekKhal on GitHub)): Text only.
- not this job: Foundry Local (Microsoft Corporation): Chat and audio transcription only; no image or speech generation.
- not this job: Core (ZMJ, Inc., doing business as Ghost, 325 9th Street, San Francisco, CA 94103, United States): No image, speech or music generation is claimed.
- not this job: GPT4All (Nomic, Inc.): Text chat only.
- not this job: HyperQwen (syv.ai (Copenhagen, CVR DK44671247)): Text only.
- not this job: Jan (Menlo Research Pte Ltd): Text models only through llama.cpp; no image, speech or music routes.
- not this job: Khoj (Khoj Inc.): Image generation and voice route to configured providers; nothing is served locally beyond chat and search.
- not this job: llama.cpp (ggml.ai, part of Hugging Face since 2026): Speech output (/v1/audio/speech), realtime transcription and multimodal input are served, but there is no image generation route.
- not this job: LM Studio (Element Labs, Inc.): Text and embeddings only; no image, speech or music routes.
- not this job: LocalGhost (LocalGhost.ai Ltd (company number 17213100, registered in England and Wales, incorporated 12 May 2026)): No generation routes.
- not this job: MLX LM (Apple Inc.): Text generation only.
- not this job: Ollama (Ollama Inc.): Text, vision input and embeddings only. No image generation, speech or music routes in the API docs.
- not this job: Open WebUI (Open WebUI Inc.): Routes image, speech and search to configured providers; serves none of them itself.
- not this job: PrivateGPT (PriBAI Technology Spain S.L.): Messages and embeddings; no image or speech routes.
- not this job: screenpipe (Negentropy Labs, Inc. (dba Screenpipe)): Not a generation server.
- not this job: Underdog (Conway Research): No generation server.
- not this job: vLLM (The Linux Foundation (vLLM is a PyTorch Foundation project)): Chat, embeddings, reranking and transcription; no image generation or speech output.

## What stops agents finishing

- password covers text routes only, image endpoints are not secured
- one GPU shared between models with no queue, so the second request waits or fails

## Around it

After: speech-to-text.self-hosted-local.

## Where this comes from

Weight 2 of 100 in Local AI: 3 of 50 accounts, all KoboldCpp or Lemonade; kept because four listings in the category advertise it

As data: https://www.anchorterminal.com/api/v1/use-cases/local-ai.multimodal-local-studio.json, or the MCP tool get_use_case with id local-ai.multimodal-local-studio. Draft, read 2026-11-12 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
