Add a target-length control for AI script generation

Users had no way to influence how long the AI-written script should
be — the system prompt's word cap was fixed at 1000 words regardless
of scenario, and the LLM's actual output length was whatever it felt
like producing within that ceiling.

Adds a Length select (1/2/5/10/15/20/30/45/60 min) next to the prompt
box. Selected minutes convert to a word target at the same 150 wpm
pace already used for the duration estimate elsewhere in the app, which
drives: the system prompt's length instruction (now phrased as a target
to reach, not just a ceiling not to exceed), the completion's max_tokens
(scaled to the target, capped at 8192), and the turn-count budget/trim
ceiling (previously a flat 50-turn cap that would have silently
truncated anything past ~20 minutes).

Long targets (30-60 min) are a best effort, not a guarantee — a single
LLM completion may not reliably produce that much coherent dialogue in
one pass, and the underlying provider's real max_tokens ceiling isn't
documented, so the longest options may come back shorter than asked.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

app.py

@@ -35,14 +35,23 @@ AVAILABLE_VOICES = list(VOICE_GENDERS.keys())
 DEFAULT_SPEAKERS = ["Cherry", "Chicago", "Janus", "Mantis"]
 
 SCRIPT_GEN_MODEL = "Qwen/Qwen2.5-Coder-32B-Instruct"
-SCRIPT_MAX_WORDS = 1000           # AI generation cap
+WORDS_PER_MINUTE = 150             # Matches the pace assumed by the client's duration estimate
+DURATION_OPTIONS_MINUTES = [1, 2, 5, 10, 15, 20, 30, 45, 60]
+MAX_COMPLETION_TOKENS = 8192       # Good-faith ceiling for a single chat_completion call; the
+                                    # underlying provider may cap lower, in which case the longest
+                                    # duration options may come back shorter than requested
 MAX_SCRIPT_WORDS = 100000         # Effectively uncapped (2026-08-14, Josh) — backend chunking has no
                                    # real ceiling; the old 20,000 cap was a leftover UI guess that
                                    # blocked genuine long-form renders below the backend's actual limit
-MAX_TURNS = 50                    # Max conversation turns
+MAX_TURNS = 250                    # Hard ceiling regardless of target length (safety valve)
 AUDIO_TTL_SECONDS = 900
 
 
+def _turns_budget_for_words(target_words: int) -> int:
+    """How many turns a script of this length plausibly needs, given full-paragraph turns."""
+    return max(6, min(MAX_TURNS, round(target_words / 55)))
+
+
 # --- Load example scripts ---
 def load_example_scripts():
     examples_dir = ROOT / "text_examples"
@@ -232,7 +241,10 @@ FORMAT RULES:
 - Use EXACTLY this format for dialogue: "Speaker N: dialogue text" where N starts at 1
 - Each turn is separated by a blank line
 - Choose the right number of speakers for the scenario (1 to 4 max)
-- Keep the total script under {max_words} words
+- LENGTH TARGET: write approximately {target_words} words total — enough dialogue to fill
+  roughly {target_minutes} minute(s) of natural spoken audio. This is a target, not just a
+  ceiling: keep the conversation developing — new angles, follow-up questions, examples,
+  pushback — rather than wrapping up early. Do not stop far short of the target.
 - Output ONLY the title and script — no stage directions, no commentary, no preamble
 
 CRITICAL — ONE SPEAKER PER TURN:
@@ -337,15 +349,22 @@ def assign_voices_by_gender(genders: dict[int, str], num_speakers: int) -> list[
     return chosen
 
 
-def generate_script_from_prompt(prompt: str) -> tuple[list[dict], int, str, list[str | None]]:
+def generate_script_from_prompt(
+    prompt: str, target_minutes: int = 2
+) -> tuple[list[dict], int, str, list[str | None]]:
     """Returns (turns, num_speakers, title, voice_selections)."""
-    system = SCRIPT_SYSTEM_PROMPT.format(max_words=SCRIPT_MAX_WORDS)
+    target_minutes = target_minutes if target_minutes in DURATION_OPTIONS_MINUTES else 2
+    target_words = target_minutes * WORDS_PER_MINUTE
+    turns_budget = _turns_budget_for_words(target_words)
+    completion_tokens = min(MAX_COMPLETION_TOKENS, int(target_words * 1.6) + 400)
+
+    system = SCRIPT_SYSTEM_PROMPT.format(target_words=target_words, target_minutes=target_minutes)
     response = llm_client.chat_completion(
         messages=[
             {"role": "system", "content": system},
             {"role": "user", "content": prompt},
         ],
-        max_tokens=4096,
+        max_tokens=completion_tokens,
         temperature=0.7,
     )
     raw = response.choices[0].message.content
@@ -367,9 +386,11 @@ def generate_script_from_prompt(prompt: str) -> tuple[list[dict], int, str, list
         for t in turns
     ]
     turns = [t for t in turns if t["text"].strip()]
-    turns = turns[:MAX_TURNS]
+    turns = turns[:turns_budget]
+    # Allow some overshoot past the target before trimming — the model runs long sometimes.
+    overshoot_ceiling = int(target_words * 1.3) + 100
     total_words = sum(len(t["text"].split()) for t in turns)
-    while total_words > MAX_SCRIPT_WORDS and turns:
+    while total_words > overshoot_ceiling and turns:
         turns.pop()
         total_words = sum(len(t["text"].split()) for t in turns)
     speaker_ids = {t["speaker"] for t in turns}
@@ -534,6 +555,12 @@ async def api_parse_script(
 
 class ScriptPromptRequest(BaseModel):
     prompt: str
+    target_minutes: int = 2
+
+
+@app.get("/api/duration-options")
+async def api_duration_options() -> list[int]:
+    return DURATION_OPTIONS_MINUTES
 
 
 @app.post("/api/generate-script")
@@ -541,10 +568,11 @@ async def api_generate_script(payload: ScriptPromptRequest) -> dict:
     prompt = (payload.prompt or "").strip()
     if not prompt:
         raise HTTPException(status_code=400, detail="Please enter a prompt.")
+    target_minutes = payload.target_minutes if payload.target_minutes in DURATION_OPTIONS_MINUTES else 2
 
     try:
         script_result, parody_lines = await asyncio.gather(
-            asyncio.to_thread(generate_script_from_prompt, prompt),
+            asyncio.to_thread(generate_script_from_prompt, prompt, target_minutes),
             asyncio.to_thread(generate_parody_story, prompt),
         )
     except Exception as e:

static/app.js

@@ -41,7 +41,7 @@ const el = {};
 [
   "runtimeStatus", "runtimeLabel", "aboutBtn", "aboutDialog", "closeAboutBtn",
   "modelSelect", "speakerStepper", "voiceRows", "cfgScale", "cfgScaleValue",
-  "scriptPrompt", "generateScriptBtn", "examplePills", "openImportBtn", "scriptGenStatus",
+  "scriptPrompt", "durationSelect", "generateScriptBtn", "examplePills", "openImportBtn", "scriptGenStatus",
   "scriptTitle", "scriptDuration", "turnsList", "addTurnBtn",
   "generateBarMeta", "generateBtn",
   "statusCard", "statusTitle", "statusDesc",
@@ -331,7 +331,7 @@ el.generateScriptBtn.addEventListener("click", async () => {
     const res = await fetch("/api/generate-script", {
       method: "POST",
       headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ prompt }),
+      body: JSON.stringify({ prompt, target_minutes: Number(el.durationSelect.value) }),
     });
     const payload = await res.json();
     if (!res.ok) throw new Error(payload.detail || "Script generation failed.");
@@ -495,10 +495,11 @@ el.generateBtn.addEventListener("click", async () => {
 
 /* ---------------- Init ---------------- */
 async function init() {
-  const [models, voices, examples] = await Promise.all([
+  const [models, voices, examples, durationOptions] = await Promise.all([
     fetch("/api/models").then((r) => r.json()),
     fetch("/api/voices").then((r) => r.json()),
     fetch("/api/examples").then((r) => r.json()),
+    fetch("/api/duration-options").then((r) => r.json()),
   ]);
   state.models = models;
   state.voices = voices;
@@ -506,6 +507,10 @@ async function init() {
   state.voiceSelections = voices.slice(0, 4).map((v) => v.name);
   while (state.voiceSelections.length < 4) state.voiceSelections.push(null);
 
+  el.durationSelect.innerHTML = durationOptions.map((m) => `<option value="${m}">${m} min</option>`).join("");
+  const defaultDuration = durationOptions.includes(2) ? 2 : durationOptions[0];
+  el.durationSelect.value = String(defaultDuration);
+
   renderSidebar();
   renderTurns();
   renderExamplePills();

static/index.html

@@ -70,6 +70,7 @@
                 rows="1"
                 placeholder="Describe a scenario — a wizard and an orc debating battle strategy..."
               ></textarea>
+              <select id="durationSelect" class="duration-select" aria-label="Target length"></select>
               <button id="generateScriptBtn" class="btn btn-ink" type="button">Write with AI</button>
             </div>
             <div class="example-chips" id="examplePills"></div>

static/styles.css

@@ -214,7 +214,19 @@ input[type="range"] { flex: 1; accent-color: var(--accent); }
   box-shadow: var(--shadow);
   margin-bottom: 14px;
 }
-.composer-row { display: flex; gap: 10px; align-items: flex-end; }
+.composer-row { display: flex; gap: 10px; align-items: center; }
+.duration-select {
+  width: auto;
+  flex-shrink: 0;
+  background: var(--field);
+  border: 1px solid var(--border-strong);
+  border-radius: var(--radius-sm);
+  padding: 10px 8px;
+  font-size: 0.82rem;
+  color: var(--ink-dim);
+  align-self: stretch;
+}
+.duration-select:focus { outline: none; border-color: var(--accent); }
 .composer textarea {
   flex: 1;
   border: 1px solid var(--border-strong);