LLAMA-TTS(1)

okay bestie, welcome to llama-tts(1) — it's the user commands manual, and no cap, we're about to make your terminal era iconic. ✨ This is part 1 of 2, so buckle up. 🎀

🌸 NAME 🌸

llama-tts - llama-tts

That's literally it. The name section said what it said. 💅

🦋 DESCRIPTION 🦋

✨ common params ✨

Okay so these are the common params, and they're giving main character energy. Let's gooo. 🚀

• -h, --help, --usage — prints usage and then dips. The classic. 🙈
• --version — shows the version and build info. slay. 💅
• -cl, --cache-list — shows the list of models in the cache. like a little model menu. 🍓
• --completion-bash — prints a source-able bash completion script for llama.cpp. so you can tab-complete your heart out. ⌨️
• -t, --threads N — number of CPU threads to use during generation (default: <num_cpus>). (env: LLAMA_ARG_THREADS)
• -tb, --threads-batch N — number of threads to use during batch and prompt processing (default: same as --threads). batch era. 💖
• -C, --cpu-mask M — CPU affinity mask: arbitrarily long hex. Complements cpu-range (default: ""). hex girl vibes. 🔮
• -Cr, --cpu-range lo-hi — range of CPUs for affinity. Complements --cpu-mask. it's giving range. 🌈
• --cpu-strict <0|1> — use strict CPU placement (default: 0). strict like your fave teacher. 😤
• --prio N — set process/thread priority: low(-1), normal(0), medium(1), high(2), realtime(3) (default: 0). prioritize yourself, bestie. 👑
• --poll <0...100> — use polling level to wait for work (0 - no polling, default: 50). poll like it's a vibe check. 📊
• -Cb, --cpu-mask-batch M — CPU affinity mask: arbitrarily long hex. Complements cpu-range-batch (default: same as --cpu-mask). batch hex, iykyk. 🔥
• -Crb, --cpu-range-batch lo-hi — ranges of CPUs for affinity. Complements --cpu-mask-batch. ranges but make it batch. 🌸
• --cpu-strict-batch <0|1> — use strict CPU placement (default: same as --cpu-strict). strict batch energy. 💀
• --prio-batch N — set process/thread priority: 0-normal, 1-medium, 2-high, 3-realtime (default: 0). priorities for the batch girlies. 👑
• --poll-batch <0|1> — use polling to wait for work (default: same as --poll). polling in batch mode, fr fr. 🧚
• -c, --ctx-size N — size of the prompt context (default: 0, 0 = loaded from model). (env: LLAMA_ARG_CTX_SIZE). context is everything. 💖
• -n, --predict, --n-predict N — number of tokens to predict (default: -1, -1 = infinity). (env: LLAMA_ARG_N_PREDICT). predict the future, no cap. 🔮
• -b, --batch-size N — logical maximum batch size (default: 2048). (env: LLAMA_ARG_BATCH). big batch energy. 🍓
• -ub, --ubatch-size N — physical maximum batch size (default: 512). (env: LLAMA_ARG_UBATCH). physical? periodt. 💅
• --keep N — number of tokens to keep from the initial prompt (default: 0, -1 = all). keep the vibes. 🎀
• --swa-full — use full-size SWA cache (default: false). more info: https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055 (env: LLAMA_ARG_SWA_FULL). SWA full, slay. ✨
• -fa, --flash-attn [on|off|auto] — set Flash Attention use ('on', 'off', or 'auto', default: 'auto'). (env: LLAMA_ARG_FLASH_ATTN). flash attention is the moment. ⚡
• -p, --prompt PROMPT — prompt to start generation with; for system message, use -sys. write your prompt, bestie. ✍️
• --perf, --no-perf — whether to enable internal libllama performance timings (default: false). (env: LLAMA_ARG_PERF). performance, but make it optional. 📈
• -f, --file FNAME — a file containing the prompt (default: none). file-based prompt, very demure. 📁
• -bf, --binary-file FNAME — binary file containing the prompt (default: none). binary file? mysterious. 🌚
• -e, --escape, --no-escape — whether to process escapes sequences (\n, \r, \t, \', \", \\) (default: true). escape sequences go brrr. 🦋
• --rope-scaling {none,linear,yarn} — RoPE frequency scaling method, defaults to linear unless specified by the model. (env: LLAMA_ARG_ROPE_SCALING_TYPE). rope scaling, it's giving options. 🪢
• --rope-scale N — RoPE context scaling factor, expands context by a factor of N. (env: LLAMA_ARG_ROPE_SCALE). scale it up. 📏
• --rope-freq-base N — RoPE base frequency, used by NTK-aware scaling (default: loaded from model). (env: LLAMA_ARG_ROPE_FREQ_BASE). frequency base, the tea. 🍵
• --rope-freq-scale N — RoPE frequency scaling factor, expands context by a factor of 1/N. (env: LLAMA_ARG_ROPE_FREQ_SCALE). divide and conquer. ➗
• --yarn-orig-ctx N — YaRN: original context size of model (default: 0 = model training context size). (env: LLAMA_ARG_YARN_ORIG_CTX). yarn, but for context. 🧶
• --yarn-ext-factor N — YaRN: extrapolation mix factor (default: -1.00, 0.0 = full interpolation). (env: LLAMA_ARG_YARN_EXT_FACTOR). mix it up. 🎛️
• --yarn-attn-factor N — YaRN: scale sqrt(t) or attention magnitude (default: -1.00). (env: LLAMA_ARG_YARN_ATTN_FACTOR). attention, hunny. 👀
• --yarn-beta-slow N — YaRN: high correction dim or alpha (default: -1.00). (env: LLAMA_ARG_YARN_BETA_SLOW). slow beta, high key. 🐢
• --yarn-beta-fast N — YaRN: low correction dim or beta (default: -1.00). (env: LLAMA_ARG_YARN_BETA_FAST). fast beta, low key. 🐇
• -kvo, --kv-offload, -nkvo, --no-kv-offload — whether to enable KV cache offloading (default: enabled). (env: LLAMA_ARG_KV_OFFLOAD). KV cache offloading, on or off. 🔛
• --repack, -nr, --no-repack — whether to enable weight repacking (default: enabled). (env: LLAMA_ARG_REPACK). repack the weight, periodt. 🎒
• --no-host — bypass host buffer allowing extra buffers to be used. (env: LLAMA_ARG_NO_HOST). no host, just buffers. 🚫
• -ctk, --cache-type-k TYPE — KV cache data type for K. allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16). (env: LLAMA_ARG_CACHE_TYPE_K). K cache type, slay. 🇰
• -ctv, --cache-type-v TYPE — KV cache data type for V. allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16). (env: LLAMA_ARG_CACHE_TYPE_V). V cache type, iconic. 🇻
• -dt, --defrag-thold N — KV cache defragmentation threshold (DEPRECATED). (env: LLAMA_ARG_DEFRAG_THOLD). not me still using this deprecated flag, iykyk. 💀
• -np, --parallel N — number of parallel sequences to decode (default: 1). (env: LLAMA_ARG_N_PARALLEL). parallel sequences, multitasking queen. 👑
• --mlock — DEPRECATED in favor of --load-mode: force system to keep model in RAM rather than swapping or compressing. (env: LLAMA_ARG_MLOCK). deprecated, so it's giving throwback. 🕰️
• --mmap, --no-mmap — DEPRECATED in favor of --load-mode: whether to memory-map model. (if mmap disabled, slower load but may reduce pageouts if not using mlock). (env: LLAMA_ARG_MMAP). also deprecated, the way that mmap works is not our problem anymore. 📼
• -dio, --direct-io, -ndio, --no-direct-io — DEPRECATED in favor of --load-mode: use DirectIO if available. (env: LLAMA_ARG_DIO). deprecated direct IO, the drama. 🎭
• -lm, --load-mode MODE — model loading mode (default: auto). (env: LLAMA_ARG_LOAD_MODE)
  • auto: mmap, unless a device does not support it
  • none: no special loading mode
  • mmap: memory-map model (if mmap disabled, slower load but may reduce pageouts if not using mlock)
  • mlock: force system to keep model in RAM rather than swapping or compressing
  • mmap+mlock: mmap + force system to keep model in RAM rather than swapping or compressing
  • dio: use DirectIO if available
• --numa TYPE — attempt optimizations that help on some NUMA systems. (env: LLAMA_ARG_NUMA)
  • distribute: spread execution evenly over all nodes
  • isolate: only spawn threads on CPUs on the node that execution started on
  • numactl: use the CPU map provided by numactl if run without this previously, it is recommended to drop the system page cache before using this. see https://github.com/ggml-org/llama.cpp/issues/1437
• -dev, --device <dev1,dev2,..> — comma-separated list of devices to use for offloading (none = don't offload). use --list-devices to see a list of available devices. (env: LLAMA_ARG_DEVICE). device list, let's go. 📱
• --list-devices — print list of available devices and exit. like a device menu. 🖥️
• -ot, --override-tensor <tensor name pattern>=<buffer type>,... — override tensor buffer type. (env: LLAMA_ARG_OVERRIDE_TENSOR). override that tensor, boss move. 💪
• -cmoe, --cpu-moe — keep all Mixture of Experts (MoE) weights in the CPU. (env: LLAMA_ARG_CPU_MOE). MoE on CPU, no cap. 🧠
• -ncmoe, --n-cpu-moe N — keep the Mixture of Experts (MoE) weights of the first N layers in the CPU. (env: LLAMA_ARG_N_CPU_MOE). first N layers, specifically. 🔢
• -ngl, --gpu-layers, --n-gpu-layers N — max. number of layers to store in VRAM, either an exact number, 'auto', or 'all' (default: auto). (env: LLAMA_ARG_N_GPU_LAYERS). GPU layers, auto or all, iconic. 🎮
• -sm, --split-mode {none,layer,row,tensor} — how to split the model across multiple GPUs, one of: (env: LLAMA_ARG_SPLIT_MODE)
  • none: use one GPU only
  • layer (default): split layers and KV across GPUs (pipelined)
  • row: split weight across GPUs by rows (parallelized)
  • tensor: split weights and KV across GPUs (parallelized, EXPERIMENTAL)
• -ts, --tensor-split N0,N1,N2,... — fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1. (env: LLAMA_ARG_TENSOR_SPLIT). tensor split, proportions, math girlie. 🧮
• -mg, --main-gpu INDEX — the GPU to use for the model (with split-mode = none), or for intermediate results and KV (with split-mode = row) (default: 0). (env: LLAMA_ARG_MAIN_GPU). main GPU, the chosen one. 👑
• -fit, --fit [on|off] — whether to adjust unset arguments to fit in device memory ('on' or 'off', default: 'on'). (env: LLAMA_ARG_FIT). fit it to the device, slay. 🧩
• -fitt, --fit-target MiB0,MiB1,MiB2,... — target margin per device for --fit, comma-separated list of values, single value is broadcast across all devices, default: 1024. (env: LLAMA_ARG_FIT_TARGET). fit target, margins, numbers. 📊
• -fitc, --fit-ctx N — minimum ctx size that can be set by --fit option, default: 4096. (env: LLAMA_ARG_FIT_CTX). minimum ctx, don't go below. 🚧
• --check-tensors — check model tensor data for invalid values (default: false). safety first, bestie, ate that. 🛡️
• --override-kv KEY=TYPE:VALUE,... — advanced option to override model metadata by key. to specify multiple overrides, either use comma-separated values. types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false,tokenizer.ggml.add_eos_token=bool:false. advanced, iykyk. 🧪
• --op-offload, --no-op-offload — whether to offload host tensor operations to device (default: true). offload those ops. 🚚
• --lora FNAME — path to LoRA adapter (use comma-separated values to load multiple adapters). LoRA adapters, load 'em up. 🎀
• --lora-scaled FNAME:SCALE,... — path to LoRA adapter with user defined scaling (format: FNAME:SCALE,...). note: use comma-separated values. scaled LoRA, fancy. 💅
• --control-vector FNAME — add a control vector. note: use comma-separated values to add multiple control vectors. control vector, the power. 🎛️
• --control-vector-scaled FNAME:SCALE,... — add a control vector with user defined scaling SCALE. note: use comma-separated values (format: FNAME:SCALE,...). scaled control vector, iconic. ✨
• --control-vector-layer-range START END — layer range to apply the control vector(s) to, start and end inclusive. layers, start to end. 📚
• -m, --model FNAME — model path to load. (env: LLAMA_ARG_MODEL). the model, the main character. 🌟
• -mu, --model-url MODEL_URL — model download url (default: unused). (env: LLAMA_ARG_MODEL_URL). download the model from a URL. 🔗
• -dr, --docker-repo [<repo>/]<model>[:quant] — ngl, the original doesn't say more, bestie, it's giving mystery. 🙈

Okay bestie, picking up exactly where we left off 🫶 — no re-intros, we're mid-list and we're moving.

So that Docker Hub model repository thing? Here's the tea: repo is optional and defaults to ai/, and quant is optional and defaults to :latest. Example: gemma3 (default: unused) (env: LLAMA_ARG_DOCKER_REPO). Slay. Moving on 💅

• -hf, -hfr, --hf-repo <user>/<model>[:quant] — your Hugging Face model repository. The quant part is optional and case-insensitive, defaults to Q4_K_M, and if Q4_K_M doesn't exist it falls back to the first file in the repo. It'll also snag mmproj automatically if it's available; to disable that, add --no-mmproj. Example: ggml-org/GLM-4.7-Flash-GGUF:Q4_K_M (default: unused) (env: LLAMA_ARG_HF_REPO). Iconic ✨
• -hff, --hf-file FILE — a specific Hugging Face model file. If you specify this one, it overrides the quant in --hf-repo (default: unused) (env: LLAMA_ARG_HF_FILE). She said "I'm the main character now" 💅
• -hft, --hf-token TOKEN — your Hugging Face access token (default: the value from the HF_TOKEN environment variable) (env: HF_TOKEN). Gated repos, we're coming for you 🔑
• --log-disable — Log disable. That's it, that's the whole vibe.
• --log-file FNAME — Log to a file (env: LLAMA_ARG_LOG_FILE) 🍓
• --log-colors [on|off|auto] — set colored logging: 'on', 'off', or 'auto', default 'auto'. And 'auto' means colors turn on when the output is going to a terminal (env: LLAMA_ARG_LOG_COLORS). Rainbow terminal, we love to see it 🌈
• -v, --verbose, --log-verbose — crank the verbosity level to infinity, i.e. log all the messages, which is genuinely useful for debugging 🔥
• --offline — offline mode: forces use of the cache and prevents network access (env: LLAMA_ARG_OFFLINE). Airplane mode but make it dev 👑
• -lv, --verbosity, --log-verbosity N — set the verbosity threshold. Any message with a higher verbosity than this gets ignored. The values:
  • 0 — generic output
  • 1 — error
  • 2 — warning
  • 3 — info
  • 4 — trace (more info)
  • 5 — debug

(default: 3) (env: LLAMA_ARG_LOG_VERBOSITY)

• --log-prefix, --no-log-prefix — enable prefix in log messages — or turn it off with the --no- twin (env: LLAMA_ARG_LOG_PREFIX) 🎀
• --log-timestamps, --no-log-timestamps — enable timestamps in log messages — or nope out with the --no- twin (env: LLAMA_ARG_LOG_TIMESTAMPS) ⏰
• --spec-draft-type-k, -ctkd, --cache-type-k-draft TYPE — the KV cache data type for K for the draft model. Allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16) (env: LLAMA_ARG_SPEC_DRAFT_CACHE_TYPE_K)
• --spec-draft-type-v, -ctvd, --cache-type-v-draft TYPE — same energy, but for V on the draft model. Allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16) (env: LLAMA_ARG_SPEC_DRAFT_CACHE_TYPE_V)

---

✨ sampling params ✨

• --samplers SAMPLERS — the samplers that will be used for generation, in order, separated by ';'. Default: penalties;dry;top_n_sigma;top_k;typ_p;top_p;min_p;xtc;temperature
• -s, --seed SEED — RNG seed (default: -1, and -1 means use a random seed). Chaos mode activated 🎲
• --sampler-seq, --sampling-seq SEQUENCE — the simplified sequence for the samplers that will be used (default: edskypmxt)
• --ignore-eos — ignore the end of stream token and keep on generating (which implies --logit-bias EOS-inf). She literally will not stop talking 😭
• --temp, --temperature N — temperature (default: 0.80)
• --top-k N — top-k sampling (default: 40, 0 = disabled) (env: LLAMA_ARG_TOP_K)
• --top-p N — top-p sampling (default: 0.95, 1.0 = disabled)
• --min-p N — min-p sampling (default: 0.05, 0.0 = disabled)
• --top-nsigma, --top-n-sigma N — top-n-sigma sampling (default: -1.00, -1.0 = disabled)
• --xtc-probability N — xtc probability (default: 0.00, 0.0 = disabled)
• --xtc-threshold N — xtc threshold (default: 0.10, 1.0 = disabled)
• --typical, --typical-p N — locally typical sampling, parameter p (default: 1.00, 1.0 = disabled)
• --repeat-last-n N — the last n tokens to consider for penalty (default: -1, 0 = disabled)
• --repeat-penalty N — penalize a repeating sequence of tokens (default: 1.05, 1.0 = disabled)
• --presence-penalty N — repeat alpha presence penalty (default: 0.00, 0.0 = disabled)
• --frequency-penalty N — repeat alpha frequency penalty (default: 0.00, 0.0 = disabled)
• --dry-multiplier N — set the DRY sampling multiplier (default: 0.00, 0.0 = disabled)
• --dry-base N — set the DRY sampling base value (default: 1.75)
• --dry-allowed-length N — set the allowed length for DRY sampling (default: 2)
• --dry-penalty-last-n N — set the DRY penalty for the last n tokens (default: 64, 0 = disable)
• --dry-sequence-breaker STRING — add a sequence breaker for DRY sampling, which clears out the default breakers ('\n', ':', '"', '*') in the process. Use "none" if you don't want any sequence breakers at all 🚫
• --adaptive-target N — adaptive-p: select tokens near this probability (valid range 0.0 to 1.0; negative = disabled) (default: -1.00). More info at https://github.com/ggml-org/llama.cpp/pull/17927
• --adaptive-decay N — adaptive-p: decay rate for target adaptation over time. Lower values are more reactive, higher values are more stable. (valid range 0.0 to 0.99) (default: 0.90)
• --dynatemp-range N — dynamic temperature range (default: 0.00, 0.0 = disabled)
• --dynatemp-exp N — dynamic temperature exponent (default: 1.00)
• --mirostat N — use Mirostat sampling. Heads up bestie: Top K, Nucleus and Locally Typical samplers are ignored if you use this. (default: 0, 0 = disabled, 1 = Mirostat, 2 = Mirostat 2.0)
• --mirostat-lr N — Mirostat learning rate, parameter eta (default: 0.10)
• --mirostat-ent N — Mirostat target entropy, parameter tau (default: 5.00)
• -l, --logit-bias TOKEN_ID(+/-)BIAS — modifies the likelihood of a token appearing in the completion. So like, --logit-bias 15043+1 to increase the likelihood of the token ' Hello', or --logit-bias 15043-1 to decrease it ✨
• --grammar GRAMMAR — a BNF-like grammar to constrain generations (peep the samples in the grammars/ dir)
• --grammar-file FNAME — file to read the grammar from
• -j, --json-schema SCHEMA — a JSON schema to constrain generations (https://json-schema.org/), e.g. {} for any JSON object. For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead
• -jf, --json-schema-file FILE — a file containing a JSON schema to constrain generations (https://json-schema.org/), e.g. {} for any JSON object. For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead
• -bs, --backend-sampling — enable backend sampling — it's experimental, so be chill about it (default: disabled) (env: LLAMA_ARG_BACKEND_SAMPLING). 🧪

---

🦋 example-specific params 🦋

• -mm, --mmproj FILE — path to a multimodal projector file. See tools/mtmd/README.md. Note: if -hf is used, this argument can be omitted (env: LLAMA_ARG_MMPROJ) 👀
• -mmu, --mmproj-url URL — URL to a multimodal projector file. See tools/mtmd/README.md (env: LLAMA_ARG_MMPROJ_URL)
• --mmproj-offload, --no-mmproj-offload — whether to enable GPU offloading for the multimodal projector (default: enabled) (env: LLAMA_ARG_MMPROJ_OFFLOAD) 🔥
• --image-min-tokens N — the minimum number of tokens each image can take, only used by vision models with dynamic resolution (default: read from model) (env: LLAMA_ARG_IMAGE_MIN_TOKENS)
• --image-max-tokens N — the maximum number of tokens each image can take, only used by vision models with dynamic resolution (default: read from model) (env: LLAMA_ARG_IMAGE_MAX_TOKENS)
• -o, --output, --output-file FNAME — output file (default: 'output.wav') 🎧
• --tts-lang FNAME — language (ISO 639-1) for audio generation. See tts/README.md for per-model usage notes
• --tts-speaker-file FNAME — speaker file path for audio generation

example usage:

llama-tts -m backbone.gguf -mm mmproj.gguf -p "text to speak" -o output.wav model -p "text to speak" -o output.wav

note: --tts-lang and --tts-speaker-file may not be supported in all models. Use -n to limit the output length. See tts/README.md for per-model usage notes 💜

---

And that's a wrap, bestie 💖 this page came to you from debian, dated September 2026, and it lives in the LLAMA-TTS(1) section of your manual — go make some audio, you absolute icon 👑✨